Large-scale Image Datasets Research Articles

Automatic segmentation of large-scale CT image datasets for detailed body composition analysis

BackgroundBody composition (BC) is an important factor in determining the risk of type 2-diabetes and cardiovascular disease. Computed tomography (CT) is a useful imaging technique for studying BC, however manual segmentation of CT images is time-consuming and subjective. The purpose of this study is to develop and evaluate fully automated segmentation techniques applicable to a 3-slice CT imaging protocol, consisting of single slices at the level of the liver, abdomen, and thigh, allowing detailed analysis of numerous tissues and organs.MethodsThe study used more than 4000 CT subjects acquired from the large-scale SCAPIS and IGT cohort to train and evaluate four convolutional neural network based architectures: ResUNET, UNET++, Ghost-UNET, and the proposed Ghost-UNET++. The segmentation techniques were developed and evaluated for automated segmentation of the liver, spleen, skeletal muscle, bone marrow, cortical bone, and various adipose tissue depots, including visceral (VAT), intraperitoneal (IPAT), retroperitoneal (RPAT), subcutaneous (SAT), deep (DSAT), and superficial SAT (SSAT), as well as intermuscular adipose tissue (IMAT). The models were trained and validated for each target using tenfold cross-validation and test sets.ResultsThe Dice scores on cross validation in SCAPIS were: ResUNET 0.964 (0.909–0.996), UNET++ 0.981 (0.927–0.996), Ghost-UNET 0.961 (0.904–0.991), and Ghost-UNET++ 0.968 (0.910–0.994). All four models showed relatively strong results, however UNET++ had the best performance overall. Ghost-UNET++ performed competitively compared to UNET++ and showed a more computationally efficient approach.ConclusionFully automated segmentation techniques can be successfully applied to a 3-slice CT imaging protocol to analyze multiple tissues and organs related to BC. The overall best performance was achieved by UNET++, against which Ghost-UNET++ showed competitive results based on a more computationally efficient approach. The use of fully automated segmentation methods can reduce analysis time and provide objective results in large-scale studies of BC.

See, caption, cluster: Large-scale image analysis using captioning and topic modeling

Owing to the widespread use of smartphones and mobile devices and the prevalence of image-sharing social network services, the amount of image data available on the Web is soaring. Various tasks, such as image classification, detection, and segmentation, use tremendous amounts of image data to train machine learning models. Using these trained models, a visual feature representation vector can be extracted from individual images and subsequently be used in several applications, such as image retrieval, object detection, and clustering. However, despite the increasing demand for such analyses, few studies have analyzed the information summarized by such image datasets, especially for extracting topics, trends, and opinions from images generated by online communities. Therefore, we propose a novel approach to image topic modeling, which accounts for visual content as well as semantic information by leveraging the image captioning model. In addition, we propose an image–caption scoring model that measures the semantic similarity between an image and its generated caption in order to filter noisy data that obstruct analysis by obscuring the semantic meaning of topics extracted from the dataset. The results show that our proposed method assists in analyzing large-scale image datasets without the need to manually check individual images. Further experimental results show that our methods are particularly beneficial for applications such as data visualization, image retrieval, and image tag recommendation in the realm of large-scale image dataset analysis.

Deep learning based multi-view stereo matching and 3D scene reconstruction from oblique aerial images

In this paper, we propose a practical three-dimensional (3D) real-scene reconstruction framework named Deep3D, which is paired with a deep learning based multi-view stereo (MVS) matching model named the adaptive multi-view aggregation matching (Ada-MVS) model, to obtain a 3D textured mesh model from multi-view oblique aerial images. Deep3D is the first deep learning based framework for 3D scene reconstruction, in which aerial triangulation and view selection are first performed on the input images, and the depth map of each image is then inferred using the pretrained Ada-MVS model. All the inferred depth maps are then fused into a dense point cloud after filtering the outliers. Finally, the 3D textured mesh is extracted from the dense 3D points as the final product. In the Ada-MVS model, a novel adaptive inter-view aggregation module is specially proposed to address the inconsistent information among oblique views and to fuse the multi-view costs into a robust cost volume. A lightweight recurrent regularization module is also designed for high-efficiency processing of high-capacity aerial images with large depth variations. Moreover, as oblique aerial image datasets are currently lacking, we built a large-scale synthetic multi-view oblique aerial image dataset (WHU-OMVS dataset) for deep learning based model training and methodology evaluation for the task of 3D scene reconstruction. The experimental results show that, firstly, the proposed Ada-MVS model has obvious advantages when used with high-capacity oblique aerial images, compared with several relevant learning-based MVS methods. Secondly, through a comprehensive comparison with popular commercial software packages and open-source solutions, it is shown that the proposed Deep3D framework outperforms all the other solutions in terms of reconstruction quality, and outperforms all the open-source solutions and some of the software packages in terms of efficiency on the WHU-OMVS dataset. Thirdly, the Deep3D framework shows a stable generalization ability and excellent performance when applied to other oblique or nadir aerial images, without any further fine-tuning. The dataset and code will be available at http://gpcv.whu.edu.cn/data.

Dynamic Parallel Pyramid Networks for Scene Recognition.

Scene recognition is considered a challenging task of image recognition, mainly due to the presence of multiscale information of global layout and local objects in a given scene. Recent convolutional neural networks (CNNs) that can learn multiscale features have achieved remarkable progress in scene recognition. They have two limitations: 1) the receptive field (RF) size is fixed even though a scene may have large-scale variations and 2) they are computing and memory intensive, partially due to the representation of multiscales. To address these limitations, we propose a lightweight dynamic scene recognition approach based on a novel architectural unit, namely, a dynamic parallel pyramid (DPP) block, that can adaptively select RF size based on multiscale information from the input regarding channel dimensions. We encode multiscale features by applying different convolutional (CONV) kernels on different input tensor channels and then dynamically merge their output using a group attention mechanism followed by channel shuffling to generate the parallel feature pyramid. DPP can be easily incorporated with existing CNNs to develop new deep models, called DPP networks (DPP-Nets). Extensive experiments on large-scale scene image datasets, Places365 standard, Places365 challenge, the Massachusetts Institute of Technology (MIT) Indoor67, and Sun397 confirmed that the proposed method provides significant performance improvement compared with current state-of-the-art (SOTA) approaches. We also verified general applicability from compelling results on lightweight models of MobileNetV2 and ShuffleNetV2 on ImageNet-1k and small object centralized benchmarks on CIFAR-10 and CIFAR-100.

Unsupervised graph-based image clustering for pretext distribution learning in IC assurance

Delayered integrated circuit (IC) image analysis is one of the most reliable approaches for the hardware assurance of ICs. Deep learning (DL) techniques have been proposed as an effective means to automate IC image analysis due to their high accuracy and efficiency, allowing for automated learning from labeled training data. These data-driven approaches typically require a large amount of comprehensive training samples that are representative of the testing dataset. Curating such suitable datasets for deep learning training is a challenging and time-consuming task, especially when working with large-scale IC image datasets. Automatically analyzing the distribution of the IC images would be an efficient method to curate ideal datasets for deep learning. In this paper, we propose an unsupervised graph-based image clustering framework for pretext distribution learning of delayered IC images. Our proposed framework consists of a pre-trained convolutional neural network (CNN) in the feature extraction stage, a graph-based feature clustering stage, and an image retrieval stage. In our experiments, we achieved a best normalized mutual information score of 0.6407 for the clustering of a benchmark dataset containing 1911 delayered IC images. Our framework has excellent ability to retrieve visually similar images from the benchmark dataset when queried with unseen image samples.

Multiplane Convex Proximal Support Vector Machine.

As an effective method for XOR problems, generalized eigenvalue proximal support vector machine (GEPSVM) recently has gained widespread attention accompanied with many variants proposed. Although these variants strengthen the classification performance to different extents, the number of fitting hyperplanes, similar to GEPSVM, for each class is still limited to just one. Intuitively, using single hyperplane seems not enough, especially for the datasets with complex feature structures. Therefore, this article mainly focuses on extending the fitting hyperplanes for each class from single one to multiple ones. However, such an extension from the original GEPSVM is not trivial even though, if possible, the elegant solution via generalized eigenvalues will also not be guaranteed. To address this issue, we first make a simple yet crucial transformation for the optimization problem of GEPSVM and then propose a novel multiplane convex proximal support vector machine (MCPSVM), where a set of hyperplanes determined by the features of the data are learned for each class. We adopt a strictly (geodesically) convex objective to characterize this optimization problem; thus, a more elegant closed-form solution is obtained, which only needs a few lines of MATLAB codes. Besides, MCPSVM is more flexible in form and can be naturally and seamlessly extended to the feature weighting learning, whereas GEPSVM and its variants can hardly straightforwardly work like this. Extensive experiments on benchmark and large-scale image datasets indicate the advantages of our MCPSVM.

Breast Tumor Segmentation in DCE-MRI With Tumor Sensitive Synthesis.

Segmenting breast tumors from dynamic contrast-enhanced magnetic resonance (DCE-MR) images is a critical step for early detection and diagnosis of breast cancer. However, variable shapes and sizes of breast tumors, as well as inhomogeneous background, make it challenging to accurately segment tumors in DCE-MR images. Therefore, in this article, we propose a novel tumor-sensitive synthesis module and demonstrate its usage after being integrated with tumor segmentation. To suppress false-positive segmentation with similar contrast enhancement characteristics to true breast tumors, our tumor-sensitive synthesis module can feedback differential loss of the true and false breast tumors. Thus, by following the tumor-sensitive synthesis module after the segmentation predictions, the false breast tumors with similar contrast enhancement characteristics to the true ones will be effectively reduced in the learned segmentation model. Moreover, the synthesis module also helps improve the boundary accuracy while inaccurate predictions near the boundary will lead to higher loss. For the evaluation, we build a very large-scale breast DCE-MR image dataset with 422 subjects from different patients, and conduct comprehensive experiments and comparisons with other algorithms to justify the effectiveness, adaptability, and robustness of our proposed method.

SFA-Net: A Selective Features Absorption Network for Object Detection in Rainy Weather Conditions.

In recent years, object detection approaches using deep convolutional neural networks (CNNs) have derived major advances in normal images. However, such success is hardly achieved with rainy images due to lack of visibility. Aiming to bridge this gap, in this article, we present a novel selective features absorption network (SFA-Net) to improve the performance of object detection not only in rainy weather conditions but also in favorable weather conditions. SFA-Net accomplishes this objective by utilizing three subnetworks, where the feature selection subnetwork is concatenated with the object detection subnetwork through the feature absorption subnetwork to form a unified model. To promote further advancement in object detection impaired by rain, we propose a large-scale rainy image dataset, named srRain, which contains both synthetic rainy images and real-world rainy images for training and testing purposes. srRain is comprised of 25 900 rainy images depicting diverse driving scenarios in the presence of rain with a total of 181 164 instances interpreting five common object categories. Experimental results display that our SFA-Net reaches the highest mean average precision (mAP) of 77.53% on a normal image set, 62.52% on a synthetic rainy image set, 37.34% on a collected natural rainy image set, and 32.86% on a published real rainy image set, surpassing current state-of-the-art object detectors and the combination of image deraining and object detection models while retaining a high speed.

A Comprehensive Implementation of Road Surface Classification for Vehicle Driving Assistance: Dataset, Models, and Deployment

The prior monitoring of the road surface conditions provides valuable information to vehicle trajectory planning and active control systems. Road surface perception with vision sensors is an emerging technology that has recently gained much attention. However, there are few reports on an exhaustive technical scheme and application practice for driving assistance. This study first creates a large-scale road surface image dataset containing one million samples with detailed road friction level, material, and unevenness level annotations. A convolutional neural network (CNN) classification model constrained by a combined loss function and adapted optimization strategies is trained on the dataset. Then we propose a decision-level fusion method based on the improved Dempster-Shafer evidence theory to enhance the robustness of the classification algorithm. Finally, the developed models are deployed on the embedded hardware platform. The top-1 accuracy for classifying road surface images reaches 92.05% by the CNN model and 97.50% after fusion. Onboard experiments illustrate the advantage and superiority of the developed technical framework, which has excellent potential in real-vehicle applications for driving assistance.

Route Planning for Emergency Evacuation Using Graph Traversal Algorithms

The automatic identification of various design elements in a floor-plan image has gained increasing attention in recent research. Emergency-evacuation applications can benefit greatly from automated floor-plan solutions, as they allow for the development of horizontal solutions instead of vertical solutions targeting a specific audience. In addition to that, current evacuation plans rely on static signs without taking into account the dynamic characteristics of each emergency case. This work aims to extract information from a floor-plan image and transform it into a graph that is used for pathfinding in an emergency evacuation. First, the basic elements of the floor-plan image, i.e., walls, rooms and doors, are identified. This is achieved using Panoptic-DeepLab, which is a state-of-the-art deep neural network for the panoptic segmentation of images, and it is available from DeepLab2, an image segmentation library. The neural network was trained using CubiCasa5K, a large-scale floor-plan image dataset containing 5000 samples, annotated into over 80 floor-plan object categories. Then, using the prediction of each pixel, a graph that shows how rooms and doors are connected is created. An application that presents this information in a user-friendly manner and provides graph editing capabilities was developed. Finally, the exits are set, and the optimal path for evacuation is calculated from each node using Dijkstra’s algorithm.

In-domain versus out-of-domain transfer learning in plankton image classification

Plankton microorganisms play a huge role in the aquatic food web. Recently, it has been proposed to use plankton as a biosensor, since they can react to even minimal perturbations of the aquatic environment with specific physiological changes, which may lead to alterations in morphology and behavior. Nowadays, the development of high-resolution in-situ automatic acquisition systems allows the research community to obtain a large amount of plankton image data. Fundamental examples are the ZooScan and Woods Hole Oceanographic Institution (WHOI) datasets, comprising up to millions of plankton images. However, obtaining unbiased annotations is expensive both in terms of time and resources, and in-situ acquired datasets generally suffer from severe imbalance, with only a few images available for several species. Transfer learning is a popular solution to these challenges, with ImageNet1K being the most-used source dataset for pre-training. On the other hand, datasets like the ZooScan and the WHOI may represent a valuable opportunity to compare out-of-domain and large-scale plankton in-domain source datasets, in terms of performance for the task at hand.In this paper, we design three transfer learning pipelines for plankton image classification, with the aim of comparing in-domain and out-of-domain transfer learning on three popular benchmark plankton datasets. The general framework consists in fine-tuning a pre-trained model on a plankton target dataset. In the first pipeline, the model is pre-trained from scratch on a large-scale plankton dataset, in the second, it is pre-trained on large-scale natural image datasets (ImageNet1K or ImageNet22K), while in the third, a two-stage fine-tuning is implemented (ImageNet rightarrow large-scale plankton dataset rightarrow target plankton dataset). Our results show that an out-of-domain ImageNet22K pre-training outperforms the plankton in-domain ones, with an average boost in test accuracy of around 6%. In the next part of this work, we adopt three ImageNet22k pre-trained Vision Transformers and one ConvNeXt, obtaining results on par (or slightly superior) with the state-of-the-art, corresponding to the usage of CNN models ensembles, with a single model. Finally, we design and test an ensemble of our Vision Transformers and the ConvNeXt, outperforming the state-of-the-art existing works on plankton image classification on the three target datasets. To support scientific community contribution and further research, our implemented code is open-source and available at https://github.com/Malga-Vision/plankton_transfer.

Parametrized pre-trained network (PPNet): A novel shape classification method using SPHARMs for MI detection

PurposeShape information contains descriptive knowledge for classification problems. A novel methodology for combining the valuable 3D and 3D + t shape information with the advantages of deep transfer learning is introduced. This method produces informative images from shape information, which are interpretable for deep CNNs previously trained using a large-scale image dataset. MethodsOur proposed structure can (1) generate 3D surfaces with optimal spatial resolution, (2) parametrize them using spherical harmonics shape descriptors (SPHARMs) with optimal frequency resolution, (3) employ registration to establish surface correspondences, and (4) convert the 3D corresponding point clouds to 2D images by spreading out the latitude and the longitude values of spherical components in the form of a flat coordinate system. The generated 2D images have the potential to feed into the pre-trained deep CNNs. The efficiency of the proposed method was assessed in myocardial infarction (MI) classification using two different datasets. ResultsFive well-known CNN architectures as fixed feature extractors were evaluated. Deep features derived from the output of convolutional and pooling layers were fed into the SVM classifier. We also explored the feasibility of adding additional information by assigning a color to each RGB channel. Subsequently, EfficinetNet-b0 reached the best performance, with an accuracy of 97.92% ± 1.47 using colored images containing additional information, respectively. ConclusionOur suggested approach by incorporating the transformed shape information and CNNs can be adopted as a robust computer-aided diagnosis system to assist clinicians in detecting various diseases such as MI.

LIVING WITH IMAGES FROM LARGE-SCALE DATA SETS: A CRITICAL PEDAGOGY FOR SCALING DOWN

The emergence of contemporary computer vision coincides with the growth and dissemination of large-scale image data sets. The grandeur of such image collections has raised fascination and concern. This article critically interrogates the assumption of scale in computer vision by asking: What can be gained by scaling down and living with images from large-scale data sets? We present results from a practice-based methodology: an ongoing exchange of individual images from data sets with selected participants. The results of this empirical inquiry help to consider how a durational engagement with such images elicits profound and variously situated meanings beyond the apparent visual content used by algorithms. We adopt the lens of critical pedagogy to untangle the role of data sets in teaching and learning, thus raising two discussion points: First, regarding how the focus on scale ignores the complexity and situatedness of images, and what it would mean for algorithms to embed more reflexive ways of seeing; Second, concerning how scaling down may support a critical literacy around data sets, raising critical consciousness around computer vision. To support the dissemination of this practice and the critical development of algorithms, we have produced a teaching plan and a tool for classroom use.

A Large-scale Synthetic Pathological Dataset for Deep Learning-enabled Segmentation of Breast Cancer

The success of training computer-vision models heavily relies on the support of large-scale, real-world images with annotations. Yet such an annotation-ready dataset is difficult to curate in pathology due to the privacy protection and excessive annotation burden. To aid in computational pathology, synthetic data generation, curation, and annotation present a cost-effective means to quickly enable data diversity that is required to boost model performance at different stages. In this study, we introduce a large-scale synthetic pathological image dataset paired with the annotation for nuclei semantic segmentation, termed as Synthetic Nuclei and annOtation Wizard (SNOW). The proposed SNOW is developed via a standardized workflow by applying the off-the-shelf image generator and nuclei annotator. The dataset contains overall 20k image tiles and 1,448,522 annotated nuclei with the CC-BY license. We show that SNOW can be used in both supervised and semi-supervised training scenarios. Extensive results suggest that synthetic-data-trained models are competitive under a variety of model training settings, expanding the scope of better using synthetic images for enhancing downstream data-driven clinical tasks.

Hash Indexing-Based Image Matching for 3D Reconstruction

Image matching is a basic task in three-dimensional reconstruction, which, in recent years, has attracted extensive attention in academic and industrial circles. However, when dealing with large-scale image datasets, these methods have low accuracy and slow speeds. To improve the effectiveness of modern image matching methods, this paper proposes an image matching method for 3D reconstruction. The proposed method can obtain high matching accuracy through hash index in a very short amount of time. The core of hash matching includes two parts: creating the hash table and hash index. The former is used to encode local feature descriptors into hash codes, and the latter is used to search candidates for query feature points. In addition, the proposed method is extremely robust to image scaling and transformation by using various verifications. A comprehensive experiment was carried out using several challenging datasets to evaluate the performance of hash matching. Experimental results show that the HashMatch presents excellent results compared to the state-of-the-art methods in both computational efficiency and matching accuracy.

Graph Flow: Cross-layer Graph Flow Distillation for Dual Efficient Medical Image Segmentation.

With the development of deep convolutional neural networks, medical image segmentation has achieved a series of breakthroughs in recent years. However, high-performance convolutional neural networks always mean numerous parameters and high computation costs, which will hinder the applications in resource-limited medical scenarios. Meanwhile, the scarceness of large-scale annotated medical image datasets further impedes the application of high-performance networks. To tackle these problems, we propose Graph Flow, a comprehensive knowledge distillation framework, for both network efficiency and annotation-efficiency medical image segmentation. Specifically, the Graph Flow Distillation transfers the essence of cross-layer variations from a well trained cumbersome teacher network to a non-trained compact student network. In addition, an unsupervised Paraphraser Module is integrated to purify the knowledge of the teacher, which is also beneficial for the training stabilization. Furthermore, we build a unified distillation framework by integrating the adversarial distillation and the vanilla logits distillation, which can further refine the final predictions of the compact network. With different teacher networks (traditional convolutional architecture or prevalent transformer architecture) and student networks, we conduct extensive experiments on four medical image datasets with different modalities (Gastric Cancer, Synapse, BUSI, and CVC-ClinicDB).We demonstrate the prominent ability of our method on these datasets, which achieves competitive performances. Moreover, we demonstrate the effectiveness of our Graph Flow through a novel semi-supervised paradigm for dual efficient medical image segmentation. Our code will be available at Graph Flow.

Convolutional neural networks for real-time wood plank detection and defect segmentation

Background: Defect detection and segmentation on product surfaces in industry has become one of the most important steps in quality control. There are many sophisticated hardware and software tools used in the industry for this purpose. The need for the real-time classification and detection of defects in industrial quality control has become a crucial requirement. Most algorithms and deep neural network architectures require expensive hardware to perform inference in real-time. This necessitates the design of architectures that are light-weight and suitable for deployment in industrial environments. Methods: In this study, we introduce a novel method for detecting wood planks on a fast-moving conveyor and using a convolutional neural network (CNN) to segment surface defects in real-time. A backbone network is trained with a large-scale image dataset. A dataset of 5000 images is created with proper annotation of wood planks and defects. In addition, a data augmentation technique is employed to enhance the accuracy of the model. Furthermore, we examine both statistical and deep learning-based approaches to identify and separate defects using the latest methods. Results: Our plank detection method achieved an impressive mean average precision of 97% and 96% of global pixel accuracy for defect segmentation. This remarkable performance is made possible by the real-time processing capabilities of our system, which can run at 30 frames per second (FPS) without sacrificing accuracy. Conclusions: The results of our study demonstrate the potential of our method not only in industrial wood processing applications but also in other industries where materials undergo similar processes of defect detection and segmentation. By utilizing our method, these industries can expect to see improved efficiency, accuracy, and overall productivity.

An Attention-Based Deep Convolutional Neural Network for Brain Tumor and Disorder Classification and Grading in Magnetic Resonance Imaging

This study proposes the integration of attention modules, feature-fusion blocks, and baseline convolutional neural networks for developing a robust multi-path network that leverages its multiple feature-extraction blocks for non-hierarchical mining of important medical image-related features. The network is evaluated using 10-fold cross-validation on large-scale magnetic resonance imaging datasets involving brain tumor classification, brain disorder classification, and dementia grading tasks. The Attention Feature Fusion VGG19 (AFF-VGG19) network demonstrates superiority against state-of-the-art networks and attains an accuracy of 0.9353 in distinguishing between three brain tumor classes, an accuracy of 0.9565 in distinguishing between Alzheimer’s and Parkinson’s diseases, and an accuracy of 0.9497 in grading cases of dementia.

UID2021: An Underwater Image Dataset for Evaluation of No-Reference Quality Assessment Metrics

Achieving subjective and objective quality assessment of underwater images is of high significance in underwater visual perception and image/video processing. However, the development of underwater image quality assessment (UIQA) is limited for the lack of publicly available underwater image datasets with human subjective scores and reliable objective UIQA metrics. To address this issue, we establish a large-scale underwater image dataset, dubbed UID2021, for evaluating no-reference (NR) UIQA metrics. The constructed dataset contains 60 multiply degraded underwater images collected from various sources, covering six common underwater scenes (i.e., bluish scene, blue-green scene, greenish scene, hazy scene, low-light scene, and turbid scene), and their corresponding 900 quality improved versions are generated by employing 15 state-of-the-art underwater image enhancement and restoration algorithms. Mean opinion scores with 52 observers for each image of UID2021 are also obtained by using the pairwise comparison sorting method. Both in-air and underwater-specific NR IQA algorithms are tested on our constructed dataset to fairly compare their performance and analyze their strengths and weaknesses. Our proposed UID2021 dataset enables ones to evaluate NR UIQA algorithms comprehensively and paves the way for further research on UIQA. The dataset is available at https://github.com/Hou-Guojia/UID2021 .

SupCAM: Chromosome cluster types identification using supervised contrastive learning with category-variant augmentation and self-margin loss.

Chromosome segmentation is a crucial analyzing task in karyotyping, a technique used in experiments to discover chromosomal abnormalities. Chromosomes often touch and occlude with each other in images, forming various chromosome clusters. The majority of chromosome segmentation methods only work on a single type of chromosome cluster. Therefore, the pre-task of chromosome segmentation, the identification of chromosome cluster types, requires more focus. Unfortunately, the previous method used for this task is limited by the small-scale chromosome cluster dataset, ChrCluster, and needs the help of large-scale natural image datasets, such as ImageNet. We realized that semantic differences between chromosomes and natural objects should not be ignored, and thus developed a novel two-step method called SupCAM, which could avoid overfitting only using ChrCluster and achieve a better performance. In the first step, we pre-trained the backbone network on ChrCluster following the supervised contrastive learning framework. We introduced two improvements to the model. One is called the category-variant image composition method, which augments samples by synthesizing valid images and proper labels. The other introduces angular margin into large-scale instance contrastive loss, namely self-margin loss, to increase the intraclass consistency and decrease interclass similarity. In the second step, we fine-tuned the network and obtained the final classification model. We validated the effectiveness of modules through massive ablation studies. Finally, SupCAM achieved an accuracy of 94.99% with the ChrCluster dataset, which outperformed the method used previously for this task. In summary, SupCAM significantly supports the chromosome cluster type identification task to achieve better automatic chromosome segmentation.