Image Categorization Research Articles

Harnessing deep reinforcement learning algorithms for image categorization: A multi algorithm approach

Image categorization is an important task in the field of Artificial Intelligence. Deep reinforcement learning (DRL) algorithms are effective for image categorization, especially in the real-time scenarios. In this study, we explore how different DRL algorithms can be used for classifying images in the field of artificial intelligence across multiple benchmark datasets. Our primary goal is to illustrate the way DRL algorithms perform comparable to conventional machine learning(ML) and deep learning(DL) methods. This usage of diverse data allows to evaluate thoroughly how this DRL algorithms will adapt and learn under different situations similar to real-world scenarios with unpredictable data. Our findings suggest that deep reinforcement learning algorithms outperform other algorithms in environments with diverse and complex input. The DRL methods achieved greater accuracies when compared with most of the machine learning and deep learning approaches at 1 Million timesteps. Among the DRL models, Recurrent Proximal Policy Optimization(RPPO) with an accuracy of 97.57% for MNIST dataset, 89% for KMNIST dataset, 89% for EMNIST dataset and Deep Q Network(DQN) on Fashion_MNIST dataset with an accuracy of 87.40% outperformed most of the deep learning and machine learning models and some achieving almost similar to the deep learning and machine learning models. This study not only demonstrates DRL’s capability to handle real-time challenges but also underscores its importance as a valuable tool for computer vision tasks, such as image categorization.

Image classification and reconstruction from low-density EEG

Recent advances in visual decoding have enabled the classification and reconstruction of perceived images from the brain. However, previous approaches have predominantly relied on stationary, costly equipment like fMRI or high-density EEG, limiting the real-world availability and applicability of such projects. Additionally, several EEG-based paradigms have utilized artifactual, rather than stimulus-related information yielding flawed classification and reconstruction results. Our goal was to reduce the cost of the decoding paradigm, while increasing its flexibility. Therefore, we investigated whether the classification of an image category and the reconstruction of the image itself is possible from the visually evoked brain activity measured by a portable, 8-channel EEG. To compensate for the low electrode count and to avoid flawed predictions, we designed a theory-guided EEG setup and created a new experiment to obtain a dataset from 9 subjects. We compared five contemporary classification models with our setup reaching an average accuracy of 34.4% for 20 image classes on hold-out test recordings. For the reconstruction, the top-performing model was used as an EEG-encoder which was combined with a pretrained latent diffusion model via double-conditioning. After fine-tuning, we reconstructed images from the test set with a 1000 trial 50-class top-1 accuracy of 35.3%. While not reaching the same performance as MRI-based paradigms on unseen stimuli, our approach greatly improved the affordability and mobility of the visual decoding technology.

A comprehensive dataset of annotated oral cavity images for diagnosis of oral cancer and oral potentially malignant disorders

ObjectivesThis study aims to address the critical gap of unavailability of publicly accessible oral cavity image datasets for developing machine learning (ML) and artificial intelligence (AI) technologies for the diagnosis and prognosis of oral cancer (OCA) and oral potentially malignant disorders (OPMD), with a particular focus on the high prevalence and delayed diagnosis in Asia. Materials and methodsFollowing ethical approval and informed written consent, images of the oral cavity were obtained from mobile phone cameras and clinical data was extracted from hospital records from patients attending to the Dental Teaching Hospital, Peradeniya, Sri Lanka. After data management and hosting, image categorization and annotations were done by clinicians using a custom-made software tool developed by the research team. ResultsA dataset comprising 3000 high-quality, anonymized images obtained from 714 patients were classified into four distinct categories: healthy, benign, OPMD, and OCA. Images were annotated with polygonal shaped oral cavity and lesion boundaries. Each image is accompanied by patient metadata, including age, sex, diagnosis, and risk factor profiles such as smoking, alcohol, and betel chewing habits. ConclusionResearchers can utilize the annotated images in the COCO format, along with the patients’ metadata, to enhance ML and AI algorithm development.

Block-based optimization for enhancing reversible watermarking using reduce difference expansion

In recent years, reversible watermarking has emerged as a promising technique that safely embeds data in digital images without compromising their originality. This method is particularly useful for sensitive images such as military, art, and medical images, where each pixel contains important information requiring authentication. Researchers have been attempting to develop this method further to increase payload capacity while maintaining visual quality and low computational complexity. In this study, we developed a reversible watermarking with block-based optimization based on Reduced Difference Expansion (RDE) applied to 3×3 pixel blocks, allowing for the embedding of 8?bit data. Based on experimental results from tests conducted on 2 common images and 3 medical images, our method could consistently achieve a payload capacity of up to 0.8924 bpp with a PSNR of 41.077 dB while maintaining good visual quality across various image categories, outperforming previous approaches.

Detection and classification of pests and diseases in rice leaf images based on deep learning algorithms

Rice is one of the important food crops worldwide, but it is susceptible to various pests and diseases during its growth process, leading to yield reduction and even death. Therefore, it is very important to detect and categorize rice pests and diseases in a timely and accurate manner. In this paper, we use the dataset from Kaggle public database, which is divided into two categories of images: "normal leaves" and "infested by pests and diseases", and configure AlexNet, Vgg and MobileNetV2 models based on Pytorch framework to detect and classify rice pests and diseases in a timely and accurate manner. MobileNetV2 model was configured based on Pytorch framework to detect and classify pests and diseases in rice leaf images. Through the changes of the loss and Accuracy curves of AlexNet, Vgg and MobileNetV2 models, we found that all three deep learning models have good training effect and can well predict whether the rice leaf images are infested by pests and diseases. The final results show that the prediction accuracy of all three sets of deep learning models reaches more than 97%, and they can detect pests and diseases on rice leaf images very accurately. This means that we can use these models to quickly and accurately detect and categorize rice pests and diseases, so that timely measures can be taken to prevent the spread of pests and diseases and improve the yield and quality of rice.

Efficient multi-stage feedback attention for diverse lesion in cancer image segmentation

In the domain of Computer-Aided Diagnosis (CAD) systems, the accurate identification of cancer lesions is paramount, given the life-threatening nature of cancer and the complexities inherent in its manifestation. This task is particularly arduous due to the often vague boundaries of cancerous regions, compounded by the presence of noise and the heterogeneity in the appearance of lesions, making precise segmentation a critical yet challenging endeavor. This study introduces an innovative, an iterative feedback mechanism tailored for the nuanced detection of cancer lesions in a variety of medical imaging modalities, offering a refining phase to adjust detection results. The core of our approach is the elimination of the need for an initial segmentation mask, a common limitation in iterative-based segmentation methods. Instead, we utilize a novel system where the feedback for refining segmentation is derived directly from the encoder–decoder architecture of our neural network model. This shift allows for more dynamic and accurate lesion identification. To further enhance the accuracy of our CAD system, we employ a multi-scale feedback attention mechanism to guide and refine predicted mask subsequent iterations. In parallel, we introduce a sophisticated weighted feedback loss function. This function synergistically combines global and iteration-specific loss considerations, thereby refining parameter estimation and improving the overall precision of the segmentation. We conducted comprehensive experiments across three distinct categories of medical imaging: colonoscopy, ultrasonography, and dermoscopic images. The experimental results demonstrate that our method not only competes favorably with but also surpasses current state-of-the-art methods in various scenarios, including both standard and challenging out-of-domain tasks. This evidences the robustness and versatility of our approach in accurately identifying cancer lesions across a spectrum of medical imaging contexts. Our source code can be found at https://github.com/dewamsa/EfficientFeedbackNetwork.

HCN: Hybrid Capsule Network for Fetal Plane Classification in Ultrasound Images

ABSTRACTClassifying fetal ultrasound images into different anatomical categories, such as the abdomen, brain, femur, thorax, and so forth can contribute to the early identification of potential anomalies or dangers during prenatal care. Ignoring major abnormalities that might lead to fetal death or permanent disability. This article proposes a novel hybrid capsule network architecture‐based method for identifying fetal ultrasound images. The proposed architecture increases the precision of fetal image categorization by combining the benefits of a capsule network with a convolutional neural network. The proposed hybrid model surpasses conventional convolutional network‐based techniques with an overall accuracy of 0.989 when tested on a publicly accessible dataset of prenatal ultrasound images. The results indicate that the proposed hybrid architecture is a promising approach for precisely and consistently classifying fetal ultrasound images, with potential uses in clinical settings.

TS-CAM: Token Semantic Coupled Attention Map for Weakly Supervised Object Localization.

Weakly supervised object localization (WSOL), which trains object localization models using solely image category annotations, remains a challenging problem. Existing approaches based on convolutional neural networks (CNNs) tend to miss full object extent while activating discriminative object parts. Based on our analysis, this is caused by CNN's intrinsic characteristics, which experiences difficulty to capture object semantics at long distances. In this article, we introduce the vision transformer to WSOL, with the aim to capture long-range semantic dependency of features by leveraging transformer's cascaded self-attention mechanism. We propose the token semantic coupled attention map (TS-CAM) method, which first decomposes class-aware semantics and then couples the semantics with attention maps for semantic-aware activation. To capture object semantics at long distances and avoid partial activation, TS-CAM performs spatial embedding by partitioning an image to a set of patch tokens. To incorporate object category information to patch tokens, TS-CAM reallocates category-related semantics to each patch token. The patch tokens are finally coupled with attention maps which are semantic-agnostic to perform semantic-aware object localization. By introducing semantic tokens to produce semantic-aware attention maps, we further explore the capability of TS-CAM for multicategory object localization. Experiments show that TS-CAM outperforms its CNN-CAM counterpart by 11.6% and 28.9% on ILSVRC and CUB-200-2011 datasets, respectively, improving the state-of-the-art with large margins. TS-CAM also demonstrates superiority for multicategory object localization on the Pascal VOC dataset. The code is available at github.com/yuanyao366/ts-cam-extension.

VICR: A Novel Software for Unbiased Video and Image Analysis in Scientific Research.

In scientific research, objectivity and unbiased data analysis is crucial for the validity and reproducibility of outcomes. This is particularly important for studies involving video or image categorization. A common approach of decreasing the bias is delegating data analysis to researchers unfamiliar with the experimental settings. However, this requires additional personnel and is prone to cognitive biases. Here we describe the Video & Image Cutter & Randomizer (VICR) software (https://github.com/kkihnphd/VICR), designed for unbiased analysis by segmenting and then randomizing videos or still images. VICR allows a single researcher to conduct and analyze studies in a blinded manner, eliminating the bias in analysis and streamlining the research process. We describe the features of the VICR software and demonstrate its capabilities using zebrafish behavior studies. To our knowledge, VICR is the only software for the randomization of video and image segments capable of eliminating bias in data analysis in a variety of research fields.

TOMGPT: Reliable Text-Only Training Approach for Cost-Effective Multi-modal Large Language Model

Multi-modal large language models (MLLMs), such as GPT-4, exhibit great comprehension capabilities on human instruction, as well as zero-shot ability on new downstream multi-modal tasks. To integrate the different modalities within a unified embedding space, previous MLLMs attempted to conduct visual instruction tuning with massive and high-quality image-text pair data, which requires substantial costs in data collection and training resources. In this article, we propose TOMGPT (Text-Only training Multi-modal GPT), a cost-effective MLLM tuned solely on easily accessible text data with much fewer resources. Along with pre-trained visual-linguistic coupled modality space (e.g., CLIP and ALIGN model), a text-only training strategy is devised to further project the aligned multi-modal latent space to that of LLM, endowing the LLM with visual comprehension capabilities in an efficient manner. Instead of enormous image-text training data required by previous MLLMs, we find that TOMGPT can be well-tuned with fewer yet diverse GPT-generated free-form text data, as we establish the semantic connection between LLM and pre-trained vision-language model. A quantitative evaluation is conducted on both MME and LVLM, which are recently released and extensively utilized MLLM benchmarks. The experiments reveal that TOMGPT achieved reliable performance compared to numerous models trained on a large amount of image-text pair data. Case studies are also presented, demonstrating TOMGPT’s broad understanding and dialogue capabilities across diverse image categories.

A hyperdimensional framework: Unveiling the interplay of RBP and GSN within CNNs for ultra-precise brain tumor classification

This study presents the RBP-CNN model, a convolutional neural network specifically designed for the precise classification of brain tumors in medical imaging. Conventional methods often encounter difficulties in extracting image noise and texture features, which has led to the incorporation of regional binary patterns (RBP) and Gray Standard Normalization (GSN) preprocessing techniques in CNN. The research addresses fundamental inquiries regarding the impact of the model on accuracy, false classifications, and efficiency. The novelty of RBP-CNN lies in its distinctive approach to extracting texture features, which involves optimizing pixel values through GSN preprocessing and generating regional binary patterns based on integral images. The objective of this research is to bridge a critical gap by providing a more accurate and efficient model for classifying brain tumors. The key findings reveal the exceptional performance of RBP-CNN, achieving a classification accuracy of 96% with a reduced false classification ratio of 7% across a dataset of 3000 samples. Comparative analyses position RBP-CNN as superior to alternative models in terms of accuracy, false classification rates, and efficiency. The structural insights and hyperparameter values of the model, as well as its application to the FigShare dataset, demonstrate its robustness and scalability. RBP-CNN emerges as an innovative and effective solution, advancing the field of medical image categorization. The findings of this study contribute a novel methodology, paving the way for future exploration in hyperspectral image applications and positioning RBP-CNN as a potential state-of-the-art tool for medical image analysis.

Spatial N-glycomics of the normal breast microenvironment reveals fucosylated and high-mannose N-glycan signatures related to BI-RADS density and ancestry.

Higher breast cancer mortality rates continue to disproportionally affect black women (BW) compared to white women (WW). This disparity is largely due to differences in tumor aggressiveness that can be related to distinct ancestry-associated breast tumor microenvironments (TMEs). Yet, characterization of the normal microenvironment (NME) in breast tissue and how they associate with breast cancer risk factors remains unknown. N-glycans, a glucose metabolism-linked post-translational modification, has not been characterized in normal breast tissue. We hypothesized that normal female breast tissue with distinct Breast Imaging and Reporting Data Systems (BI-RADS) categories have unique microenvironments based on N-glycan signatures that varies with genetic ancestries. Profiles of N-glycans were characterized in normal breast tissue from BW (n=20) and WW (n=20) at risk for breast cancer using matrix assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI). A total of 176 N-glycans (32 core-fucosylated and 144 noncore-fucosylated) were identified in the NME. We found that certain core-fucosylated, outer-arm fucosylated and high-mannose N-glycan structures had specific intensity patterns and histological distributions in the breast NME dependent on BI-RADS densities and ancestry. Normal breast tissue from BW, and not WW, with heterogeneously dense breast densities followed high-mannose patterns as seen in invasive ductal and lobular carcinomas. Lastly, lifestyles factors (e.g. age, menopausal status, Gail score, BMI, BI-RADS) differentially associated with fucosylated and high-mannose N-glycans based on ancestry. This study aims to decipher the molecular signatures in the breast NME from distinct ancestries towards improving the overall disparities in breast cancer burden.

SegLD: Achieving universal, zero-shot and open-vocabulary segmentation through multimodal fusion via latent diffusion processes

Open-vocabulary learning can identify categories marked during training (seen categories) and generalize to categories not annotated in the training set (unseen categories). It could theoretically extend segmentation systems to more universal applications. However, current open-vocabulary segmentation frameworks are primarily suited for specific tasks or require retraining according to the task, and they significantly underperform in inferring seen categories compared to fully supervised frameworks. Therefore, we introduce a universal open-vocabulary segmentation framework based on the latent diffusion process (SegLD), which requires only a single training session on a panoptic dataset to achieve inference across all open-vocabulary segmentation tasks, and reaches SOTA segmentation performance for both seen and unseen categories in every task. Specifically, SegLD comprises two stages: in the first stage, we deploy two parallel latent diffusion processes to deeply fuse the text (image caption or category labels) and image information, further aggregating the multi-scale features output from both latent diffusion processes on a scale basis. In the second stage, we introduce text queries, text list queries, and task queries, facilitating the learning of inter-category and inter-task differences through the computation of contrastive losses between them. Text queries are then further fed into a Transformer Decoder to obtain category-agnostic segmentation masks. Then we establish classification loss functions for the type of text input during training, whether image captions or category labels, to help assign a category label from the open vocabulary to each predicted binary mask. Experimental results show that, with just a single training session, SegLD significantly outperforms other contemporary SOTA fully supervised segmentation frameworks and open-vocabulary segmentation frameworks across almost all evaluation metrics for both known and unknown categories on the ADE20K, Cityscapes, and COCO datasets. This highlights SegLD’s capability as a universal segmentation framework, with the potential to replace other segmentation frameworks and adapt to various segmentation domains. The project link for SegLD is https://zht-segld.github.io/.

Optimizing Object Classification in Robotic Perception Environments Exploring Late Fusion Strategies

Robotic perception systems often include approaches that can extract valuable features or information from the studied dataset. These methods often involve the application of deep learning approaches, such as convolutional neural networks (CNNs), for processing of images, as well as the incorporation of 3D data. The notion of image categorization is well delineated via the use of networks that include convolutional networks. However, some network topologies exhibit a substantial scope and need significant amounts of time and memory resources. On the other hand, the neural networks FlowNet3D and PointFlowNet have the capability to accurately predict scene flow. Specifically, these networks are capable of estimating the three-dimensional movements of point clouds (PCs) within a dynamic environment. When using PCs in robotic applications, it is crucial to examine the robustness of accurately recognizing the points that belong to the object. This article examines the use of robotic perception systems inside autonomous vehicles and the inherent difficulties linked to the analysis and processing of information obtained from diverse sensors. The researchers put out a late fusion methodology that integrates the results of many classifiers in order to enhance the accuracy of categorization. Additionally, the authors propose a weighted fusion technique that incorporates the proximity to objects as a significant factor. The findings indicate that the fusion methods described in this study exhibit superior performance compared to both single modality classification and classic fusion strategies.

Holdout based blending approaches for improved satellite image classification

An essential component of remote sensing, image analysis, and pattern recognition is image categorization. The classification of land use using remotely sensed data creates a map-like representation as the final form of the investigation. With its ability to effectively categorize satellite images, machine learning (ML) algorithms have gained significant traction in a number of fields, including land-use planning, disaster response, and natural resource management. Ensemble learning is also a widely used technique for enhancing the precision of satellite image categorization, which combines multiple models to get more precise predictions. Holdout is an ensemble technique, where multiple ML algorithms are used for training on the same dataset. The primary goal of this study is to create a holdout model for classifying satellite images. Initially, this study explores the usage of ML algorithms namely support vector machines (SVM), k-nearest neighbor (KNN), decision trees (DT), gradient boosting classifier (GBC), histogram-based GBC (HGBC), random forest classifier (RF), bagging classifier (BC), XGBoost classifier for classifying satellite images. Later, GBC, HGBC, RF, BC, and XGBoost are combined to build a stacking model. The bagging ensemble model outperforms all other methods and reaches an accuracy of 88.90%. Finally, blending models with holdout approach were developed and achieved accuracy of 93.70%, 94.14%, and 93.87% which outperformed all previous algorithms.

Opportunity cost of oncology subspecialist opinion.

e13578 Background: The adoption of precision oncology has resulted in a drastic rise in the cost of cancer care in the United States. Payers, providers, and professional organizations are implementing strategies to curb this rising cost. These efforts are mainly focused on the providers patterns of care, such as prescribing behavior and risk assessment for needs such as ER visits and hospitalizations. Most efforts are cross-cutting and intended to work across different tumor types. Yet, with the precision nature of cancer care these cross-cutting efforts have not succeeded. AccessHope (AH) provides an independent subspecialist review of the management cancer as a covered benefit through the employer (Accountable Precision Oncology: APO). The aim of this service is optimizing care delivery to the current and evolving paradigms for each cancer type and sub-type. Here we describe overutilization of treatments and assessments of APO. Methods: AH database was accessed for de-identified cases (with contractual de-identification rights) for this analysis. Cases were completed 1/1/2022 – 12/31/2023. Descriptive statistics was used to describe the observed patterns identified. Results: Among 3644 APOs, 428 (11.7%) had evidence of overutilization. Gastrointestinal, hematological, and thoracic cancers represented 75% of the over utilization cases (Table). Overutilization was categorized by systemic therapy, radiation therapy, invasive procedures, imaging, and testing. Systemic therapy was the most common category (67.8%) and included non-indication use and extended duration of use. This was followed by imaging (22.2%). Systemic agents overutilized varied by cancer site. Bevacizumab and oxaliplatin were commonly overutilized in gastrointestinal cancers; rituximab, imatinib, and dasatinib in hematological malignancies; and check point inhibitors (pembrolizumab, atezolizumab, and durvalumab) in thoracic malignancies. In the imaging category, PET scan overutilization was observed across different tumor types (14.7% of the total volume). We also observed overutilization of molecular testing with ctDNA (Signatera) being at the top of the frequently overutilized tests in cases with no data to support a role for it. Conclusions: This patient centered approach identified areas of overutilization in evaluation and treatment of patients through the expert review process. The reviewer’s depth of knowledge about reported data and limitation of treatments and evaluations is the key to this process. Further analyses of patterns of care and categorization of the overutilization patterns is underway. [Table: see text]

A deep learning-based system for accurate diagnosis of pelvic bone tumors

Bone image analysis and categorizing bone cancers have both seen advancements thanks to deep learning (DL), more notably convolution neural networks (CNN). This study suggests a brand-new CNN-based methodology for categorizing pelvic bone tumors specifically. This work aims to create a pelvic bone computed tomography (CT) image categorization system based on deep learning. The proposed technique uses a convolutional neural network (CNN) architecture to automatically extract information from the CT images and classify them into distinct categories of tumors. A total of 178 3D CT pictures was discovered and added retroactively. DenseNet created the image-based model with Adam optimizer and cross entropy loss. The suggested system's accuracy is assessed using a variety of performance indicators, including sensitivity, specificity, and F1-score. As demonstrated by the experiment findings, the suggested deep learning based classification system has a high degree of accuracy (94%), making it useful for the diagnosis and treatment of pelvic bone tumors. Our promising results might hasten the use of DL-assisted CT diagnosis for pelvic bone tumors in the future.

Deep learning-based differentiation of ventricular septal defect from tetralogy of Fallot in fetal echocardiography images.

Congenital heart disease (CHD) seriously affects children's health and quality of life, and early detection of CHD can reduce its impact on children's health. Tetralogy of Fallot (TOF) and ventricular septal defect (VSD) are two types of CHD that have similarities in echocardiography. However, TOF has worse diagnosis and higher morality than VSD. Accurate differentiation between VSD and TOF is highly important for administrative property treatment and improving affected factors' diagnoses. TOF and VSD were differentiated using convolutional neural network (CNN) models that classified fetal echocardiography images. We collected 105 fetal echocardiography images of TOF and 96 images of VSD. Four CNN models, namely, VGG19, ResNet50, NTS-Net, and the weakly supervised data augmentation network (WSDAN), were used to differentiate the two congenital heart diseases. The performance of these four models was compared based on sensitivity, accuracy, specificity, and AUC. VGG19 and ResNet50 performed similarly, with AUCs of 0.799 and 0.802, respectively. A superior performance was observed with NTS-Net and WSDAN specific for fine-grained image categorization tasks, with AUCs of 0.823 and 0.873, respectively. WSDAN had the best performance among all models tested. WSDAN exhibited the best performance in differentiating between TOF and VSD and is worthy of further clinical popularization.

Food Image Classification and Calorie Prediction for Dietary Analysis

Abstract: This paper presents a unique method to classify food items and estimate the calorie content based on photos by combining Convolutional Neural Networks (CNNs) with image processing techniques. To improve the quality of the dataset, we first curate a wide range of food photographs from different presentation styles and cuisines. We do this by applying preprocessing techniques including image segmentation and feature extraction. Next, using the CNN's capacity to extract hierarchical features from raw pixel data, a custom deep CNN architecture is trained on this dataset to efficiently classify diverse food items and achieve high accuracy in differentiating between different dishes. Furthermore, we tackle the problem of calorie prediction by employing regression models that include features extracted from the CNN, which allows one to forecast the calorie content of a food item based on its visual attributes. Our technique intends to enable users to make more educated nutritional decisions and better control their caloric intake by fusing image categorization with calorie prediction. The suggested approach shows encouraging results in terms of food item classification accuracy and precision in calorie estimate. Its possible uses include promoting better eating practices and assisting individuals, dietitians, and the food industry with their dietary monitoring. In addition, our technology gives customers access to weekly calorie intake information, which can help them avoid obesity-related illnesses like diabetes.

Temporal dynamics of short-term neural adaptation across human visual cortex.

Neural responses in visual cortex adapt to prolonged and repeated stimuli. While adaptation occurs across the visual cortex, it is unclear how adaptation patterns and computational mechanisms differ across the visual hierarchy. Here we characterize two signatures of short-term neural adaptation in time-varying intracranial electroencephalography (iEEG) data collected while participants viewed naturalistic image categories varying in duration and repetition interval. Ventral- and lateral-occipitotemporal cortex exhibit slower and prolonged adaptation to single stimuli and slower recovery from adaptation to repeated stimuli compared to V1-V3. For category-selective electrodes, recovery from adaptation is slower for preferred than non-preferred stimuli. To model neural adaptation we augment our delayed divisive normalization (DN) model by scaling the input strength as a function of stimulus category, enabling the model to accurately predict neural responses across multiple image categories. The model fits suggest that differences in adaptation patterns arise from slower normalization dynamics in higher visual areas interacting with differences in input strength resulting from category selectivity. Our results reveal systematic differences in temporal adaptation of neural population responses between lower and higher visual brain areas and show that a single computational model of history-dependent normalization dynamics, fit with area-specific parameters, accounts for these differences.