Composite Image Research Articles

Automatic detection of punctate white matter lesions in infants using deep learning of composite images from two cases

Punctate white matter lesions (PWMLs) in infants may be related to neurodevelopmental outcomes based on the location or number of lesions. This study aimed to assess the automatic detectability of PWMLs in infants on deep learning using composite images created from several cases. To create the initial composite images, magnetic resonance (MR) images of two infants with the most PWMLs were used; their PWMLs were extracted and pasted onto MR images of infants without abnormality, creating many composite PWML images. Deep learning models based on a convolutional neural network, You Only Look Once v3 (YOLOv3), were constructed using the training set of 600, 1200, 2400, and 3600 composite images. As a result, a threshold of detection probability of 20% and 30% for all deep learning model sets yielded a relatively high sensitivity for automatic PWML detection (0.908–0.957). Although relatively high false-positive detections occurred with the lower threshold of detection probability, primarily, in the partial volume of the cerebral cortex (≥ 85.8%), those can be easily distinguished from the white matter lesions. Relatively highly sensitive automatic detection of PWMLs was achieved by creating composite images from two cases using deep learning.

Knee cartilage injuries in athletes

Acute and chronic cartilage injuries are often encountered in professional and recreational athletes. They can compromise the athlete's performance and career and are considered apotential risk factor for early joint degeneration. Incidence of cartilage injury in athletes, understanding of cartilage composition, injury mechanism and suitable diagnostic imaging are summarized and established therapeutic procedures, postoperative imaging including detection of relevant complications and assessment of reasonable indications for follow-up examinations are described. Original research and review articles were analyzed. Cartilage injury can mimic meniscal or ligamentous injury and cannot be ruled out by clinical examination alone. Magnetic resonance imaging (MRI) is the method of choice to (1)detect (sensitivity 87-93%, specificity 94-99%) and grade cartilage lesions to facilitate choice of therapy and (2)to exclude concomitant injuries that require treatment to improve the prognosis of the chosen cartilage therapy. Postoperatively MRI allows noninvasive assessment of the repaired cartilage tissue and is an appropriate method to detect therapeutically relevant complications. Knowledge of mechanisms and appearance of cartilage injuries, current cartilage repair techniques and their imaging is crucial for the medical care of athletes.

Developing a new red band–SEVI–blue band (RSB) enhancement method for recognition the extra-high-voltage transmission line corridor in green mountains

ABSTRACT Monitoring the extra-high-voltage transmission line corridor (EHVTLC) in mountains is critical for safe smart-grid operation. However, the transmission lines are so narrow that they are difficult to recognize using multispectral satellite images with a spatial resolution of 10 m. In this study, we developed a new method using the red band–shadow-eliminated vegetation index (SEVI)–blue band (RSB) composite image to enhance the EHVTLC in green mountains (named RSB-enhancement method). Using this method, the EHVTLC becomes evident in the false-color synthesis of the RSB composite of the Sentinel-2 image. Then, we recognized and extracted approximately 342.45 km of the EHVTLC in a mountainous region of Fuzhou City, China, including a 46.73 km three-parallel-lane segment of 1000 kV and a 295.72 km two-parallel-lane segment of 500 kV. Spatial analysis shows that the SEVI mean difference between the EHVTLC and the buffer zone reaches approximately 10%, and three landslides and 2.66 km2 soil erosion reside in the buffer zone which area is approximately 73.67 km2. Finally, the RSB-enhancement method can be used in other satellite images with spatial resolutions of greater than 10 m for enhancement and recognition the transmission line corridors in green mountains.

Imaging-based assessment of fatty acid composition in human bone marrow adipose tissue at 7T: Method comparison and in vivo feasibility.

To demonstrate the feasibility and accuracy of chemical shift-encoded imaging of the fatty acid composition (FAC) of human bone marrow adipose tissue at 7T, and to determine suitable image-acquisition parameters using simulations. The noise performance of FAC estimation was investigated using simulations with a range of inter-echo time, and accuracy was assessed using a phantom experiment. Furthermore, one knee of 8 knee-healthy subjects (ages 35-54 years) was imaged, and the fractions of saturated fatty acids (SFA) and polyunsaturated fatty acids (PUFA) were mapped. Values were compared between reconstruction methods, and between anatomical regions. Based on simulations, ΔTE=0.6ms was chosen. The phantom experiment demonstrated high accuracy of especially SFA using a constrained reconstruction model (slope=1.1, average bias=-0.2%). The lowest accuracy was seen for PUFA using a free model (slope=2.0, average bias=9.0%). For in vivo images, the constrained model resulted in lower intersubject variation compared with the free model (e.g., in the femoral shaft, the SFA percent-point range was within 1.0% [vs. 3.0%]). Furthermore, significant regional FAC differences were detected. For example, using the constrained approach, the femoral SFA in the medial condyle was lower compared with the shaft (median [range]: 27.9% [27.1%, 28.4%] vs. 32.5% [31.8%, 32.8%]). Bone marrow adipose tissue FAC quantification using chemical-shift encoding is feasible at 7T. Both the noise performance and accuracy of the technique are superior using a constrained signal model.

Identification of Fault Zone in Bali Using GGMPlus Gravity and Alos-2 Palsar-2 Data

The local active fault in Bali has a small magnitude (M<5) but has destructive potential because it is very close to residential areas. Mapping the fault area on Bali is needed to identify the parameters of faults. This study used gravity data from GGMplus, topographic data from DEMNAS, and lineaments using ALOS-2 PALSAR-2 data. Validation and interpretation using the geological map of Bali and seismicity data. We interpret the subsurface using the gravity derivative method to identify the type of fault movement. Identify fault locations using lineament extraction from SAR data processed by directional filters. The composite image red-green-blue (RGB) for HH, HV, and VV polarization was used for automatic lineament extraction and then corrected manually. The results of the gravity method succeeded in identifying 29 of the 30 faults from the geological map of the Bali sheet and a new spot from PALSAR-2. Bali land has 12 thrust faults, 11 strike-slip faults and six normal faults. The image of PALSAR-2 (L band) has succeeded in making a fault lineament map for the Bali region. The lineament extraction results from PALSAR-2 obtained four new faults (Pesanggaran, Sepang, Tegal Badeng, and Banyuwedang), while four faults were not identified (Tampaksiring Fault, Plaga, Mambal, and Munduk-Rajasa). NE-SW dominates the strike directions, and the dip angles are 45-80 degrees. We propose 30 faults in Bali, including 26 defects from geological maps with changes in length and location shift and four new marks extracted from automatic lineament.Keywords: Remote Sensing, Earthquake, Derivative Gravity, Lineament, SAR

Effect of strain rate on fractography texture descriptor of AA6063/(Si3N4)x/(Cu(NO3)2)y (x=12%, y = 2–6%)composite after multiple ECAP passes: second order statistical texture analysis conjunction with regression analysis

The tensile strength of the ECAP processed composite of AA6063/(Si3N4)x/(Cu(NO3)2)y (x=12%, y=2-6%) at various strain rates from 0.0001 to 1s-1 with an increment of 10 was examined in relation to the seven-fracture image texture descriptor through void features such as void size and distribution. In order to measure the matrices of voids on the fracture surface created through the coalescence of void results to enhance the material hardening as strain rate increases, the second order statistical texture analysis has been used. With respect to the rising strain rate, it was observed that the composite fracture image feature parameter correlation, homogeneity improved by 6 %, and 20.74% for AA6063/12%Si3N4/6%Cu(NO3)2 after I passes as compared to AA6063/12% Si3N4 after III pass. The contrast was reduced by around 58.643% when reinforcement went from a single reinforcement of 12% Si3N4 with I Pass to a composite of two reinforcements (Si3N4, Cu(NO3)2) that had 6% Cu(NO3)2 with I Pass. After three passes on AA6063/12%Si3N4/2%Cu(NO3)2, the maximum entropy was seen due to the material's increased plastic deformation, which led to an uneven particle distribution. As strain rate and copper nitrate percentage grew from 0 to 6 percent, the tensile strength of the composite improved by 34.69%. To assess the importance of fracture image features in relation to tensile strength with respect to strain rate, a regression model was developed. In order to get information about the fracture studies from fractography, the contrast with varying strain rate was nevertheless the aspect, which had the greatest influence.

Issue Information

New PhytologistVolume 237, Issue 5 p. 1475-1478 Cover and Issue InformationFree Access Issue Information First published: 02 February 2023 https://doi.org/10.1111/nph.18233AboutPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Graphical Abstract Cover Legend Composite image made with the preview of the 3D structure and 3D coloured texture of a flower 3D model reconstructed using photogrammetry. The flower belongs to Rhytidophyllum auriculatum × vernicosum (Gesneriaceae). Image courtesy of Marion Leménager (Leménager et al., pp. 1922–1933). Volume237, Issue5March 2023Pages 1475-1478 RelatedInformation

Tribological study of aluminium-B4C/Gr composite-An experimental study

Being one of the lightweight material and easily available materials aluminium has a lot of applications in various fields of automobile, electronics, and aircraft. In this study, the tribological study of aluminium-based composite is being done for the wear characteristics. A comparison is being made for the wear study of the Al 6063-based composite with the base alloy. Stir casting is used to form the composite. Microscopic images confirm the presence of Boron Carbide (B4C) and Graphite (Gr.) particles in the composite. Tribological experiments are being performed to check for the outcome of B4C and Gr.reinforcementparticles on the composite wear rate and SEM images of the worn surfaces depict that the insite presence of Gr. particles reduces the wear characteristics and improves the surface quality forthe worn composite as compared to the base alloy.

MRI image synthesis for fluid-attenuated inversion recovery and diffusion-weighted images with deep learning.

This study aims to synthesize fluid-attenuated inversion recovery (FLAIR) and diffusion-weighted images (DWI) with a deep conditional adversarial network from T1- and T2-weighted magnetic resonance imaging (MRI) images. A total of 1980 images of 102 patients were split into two datasets: 1470 (68 patients) in a training set and 510 (34 patients) in a test set. The prediction framework was based on a convolutional neural network with a generator and discriminator. T1-weighted, T2-weighted, and composite images were used as inputs. The digital imaging and communications in medicine (DICOM) images were converted to 8-bit red-green-blue images. The red and blue channels of the composite images were assigned to 8-bit grayscale pixel values in T1-weighted images, and the green channel was assigned to those in T2-weighted images. The prediction FLAIR and DWI images were of the same objects as the inputs. For the results, the prediction model with composite MRI input images in the DWI image showed the smallest relative mean absolute error (rMAE) and largest mutual information (MI), and that in the FLAIR image showed the largest relative mean-square error (rMSE), relative root-mean-square error (rRMSE), and peak signal-to-noise ratio (PSNR). For the FLAIR image, the prediction model with the T2-weighted MRI input images generated more accurate synthesis results than that with the T1-weighted inputs. The proposed image synthesis framework can improve the versatility and quality of multi-contrast MRI without extra scans. The composite input MRI image contributes to synthesizing the multi-contrast MRI image efficiently.

Shaping Ability of Protaper Gold and Protaper Ultimate in Simulated Root Canals: A Comparative Study

This study aimed to compare the shaping ability of ProTaper Gold (PTG) and the newly developed ProTaper Ultimate (PTU) systems in artificial S-shaped canals. Both files are manufactured via gold thermal treatment methods and have variable taper along their working part. Thirty-Two S-shaped canals in clear resin blocks were prepared up to an apical size of 25 using PTU and PTG instruments (n = 16 canals/group). Resin blocks were filled with ink and photographed before & after instrumentation to develop composite images by superimposition. The amount of removed resin was measured perpendicularly to the canal surface in 18 points, in addition to the canal transportation and centering ratio. Statistical analysis was performed by using Manne-Whitney U-test (α = 0.05). No broken files or aberrations were recorded during instrumentation. PTG significantly removed a higher amount of resin at 4, 6, 8, and 9 mm from the apex, but at 1 mm PTU significantly cut a higher amount (p < 0.05). Concerning transportation & centering ability, both files deviated from the center and at 3 mm PTU yielded significantly higher deviation, while at 4 mm PTG recorded a higher value (p < 0.05). In conclusion, PTU produced more conservative preparation than PTG especially in the coronal third, while for maintaining the original canal curve, PTG caused lesser deviation at the beginning of the apical curvature.

Mechanically gated formation of donor-acceptor Stenhouse adducts enabling mechanochemical multicolour soft lithography.

Stress-sensitive molecules called mechanophores undergo productive chemical transformations in response to mechanical force. A variety of mechanochromic mechanophores, which change colour in response to stress, have been developed, but modulating the properties of the dyes generally requires the independent preparation of discrete derivatives. Here we introduce a mechanophore platform enabling mechanically gated multicolour chromogenic reactivity. The mechanophore is based on an activated furan precursor to donor-acceptor Stenhouse adducts (DASAs) masked as a hetero-Diels-Alder adduct. Mechanochemical activation of the mechanophore unveils the DASA precursor, and subsequent reaction with a secondary amine generates an intensely coloured DASA. Critically, the properties of the DASA are controlled by the amine, and thus a single mechanophore can be differentiated post-activation to produce a wide range of functionally diverse DASAs. We highlight this system by establishing the concept of mechanochemical multicolour soft lithography whereby a complex multicolour composite image is printed into a mechanochemically active elastomer through an iterative process of localized compression followed by reaction with different amines.

REMOTE SENSING METHODS TO DETECT AND IDENTIFY CLAY-MADE HOUSES

Abstract. One of the most important issues in urban management is the existence of clay-made houses in or around cities. By the hypothesis that these phenomena are spectrally and spatially behaviour differently from other land uses and land covers. This study attempts to detect and identify clay-made houses through different image processing methods. In this regard, Sentinel 1, 2, and Landsat 8 OLI thermal images of Yazd city have been analysed. After pre-processing, a number of image processing techniques, including optical and radar image composite generation, band ratio, image fusion, image filtering, and image classification (i.e. MLC and ANN), have been used. The result shows that RGB from a combination of Red=radar band C, Green=sentinel band 8, and Blue=sentinel band 4 are the best for visual interpretation of clay-made houses. Also, the most suitable band ratio is the ratio corresponding to radar C and NIR, the Red and Green bands, and the best fusion is derived with Sentinel 1 (band C) and Sentinel 2 (bands 4 and 8) images through the HSV algorithm, kappa coefficients of the agreement for MLC and ANN classifiers are 0.89 and 0.29 respectively. It is concluded that remote sensing image analysis can effectively be used to detect and identify clay-made houses.

Selective Schiff base formation via gas-phase ion/ion reactions to enable differentiation of isobaric lipids in imaging mass spectrometry.

The separation and identification of lipids in complex mixtures are critical to deciphering their cellular functions. Failure to resolve isobaric compounds (e.g., via high mass resolution or tandem mass spectrometry) can result in incorrect identifications in mass spectrometry experiments. In imaging mass spectrometry, unresolved peaks can also result in composite images of multiple compounds, giving inaccurate depictions of molecular distributions. Gas-phase ion/ion reactions can be used to selectively react with specific chemical functional groups on a target analyte, thereby extracting it from a complex mixture and shifting its m/z value to an unobstructed region of the mass range. Herein, we use selective Schiff base formation via a novel charge inversion ion/ion reaction to purify phosphatidylserines from other isobaric (i.e., same nominal mass) lipids and reveal their singular distributions in imaging mass spectrometry. The selective Schiff base formation between singly deprotonated phosphatidylserine (PS) lipid anions and doubly charged N,N,N',N'-tetramethyl-N,N'-bis(6-oxohexyl)hexane-1,6-diaminium (TMODA) cations is performed using a modified commercial dual source hybrid Fourier transform ion cyclotron resonance (FTICR) mass spectrometer. This process is demonstrated using the isobaric lipids [PS 40:6 - H]- (m/z 834.528) and [SHexCer d38:1 - H]- (m/z 834.576), which produces [PS 40:6 + TMODA - H - H2O]+ (m/z 1186.879), and [SHexCer d38:1 + TMODA-H]+ (m/z 1204.938) product ions following the gas-phase charge inversion reaction. These product ions differ by roughly 18Da in mass and are easily separated by low mass resolution analysis, while the isobaric precursor ions require roughly 45,000 mass resolving power (full-width at half maximum) to separate. Imaging mass spectrometry using targeted gas-phase ion/ion reactions shows distinct spatial distributions for the separated lipid product ions relative to the composite images of the unseparated precursor ions.

Spatial-invariant convolutional neural network for photographic composition prediction and automatic correction

“Composition” determines the vividness of the image and its narrative power. Current research on image aesthetics implicitly considers simple composition rules, but no reliable composition classification and image optimization method explicitly considers composition rules. The existing composition classification models are not suitable for snapshots. We propose a composition classification model based on spatial-invariant convolutional neural networks (RSTN) with translation invariance and rotation invariance. It enhances the generalization of the model for snapshots or skewed images. Ultimately, the accuracy of the RSTN model improved by 3% over the Baseline to 90.8762%, and the rotation consistency improved by 16.015%. Furthermore, we classify images into three categories based on their sensitivity to editing: skew-sensitive, translation-sensitive, and non-space-sensitive. We design a set of composition optimization strategies for each composition that can effectively adjust the composition to beautify the image.

Automatic Shadow Generation via Exposure Fusion

Shadow generation aims to generate a plausible shadow for the inserted foreground object in a composite image. Besides the composite image and the associated mask of the inserted foreground object, existing methods also require a mask of all background objects as well as their shadows as an auxiliary input, which is laborious in practical applications. Meanwhile, most existing methods use a linear illumination transformation to darken the shadow region, which is prone to produce unrealistic shadows especially when background illumination is complex. To address these problems, this paper proposes an automatic shadow generation method, which avoids the laborious acquisition of the background object masks while harmonizing the shadow region to achieve plausible shadow effects. Specifically, to implicitly exploit background illumination to infer the shadow shape of the inserted foreground object, we first propose a Hierarchy Attention U-Net (HAU-Net) to sequentially build global interactions between the foreground object and background across spatial and channel dimensions. Since the spatial-variant property of the shadow, we formulate shadow harmonization as an exposure fusion problem and propose an Illumination-Aware Fusion Network (IFNet), which uses an improved illumination model with a double linear transformation to produce multiple under-exposure images of the shadow region. IFNet then learns pixel-wise fusion kernels that consider the local smoothness of the shadow to fuse the composite image with these under-exposure images to generate the realistic shadow of the foreground object. Extensive experiments on the DESOBA and Shadow-AR datasets demonstrate that our method achieves state-of-the-art performance for shadow generation on both the BOS and BOS-free test images.

From Global to Local: Multi-Patch and Multi-Scale Contrastive Similarity Learning for Unsupervised Defocus Blur Detection.

Defocus blur detection (DBD), which aims to detect out-of-focus or in-focus pixels from a single image, has been widely applied to many vision tasks. To remove the limitation on the abundant pixel-level manual annotations, unsupervised DBD has attracted much attention in recent years. In this paper, a novel deep network named Multi-patch and Multi-scale Contrastive Similarity (M2CS) learning is proposed for unsupervised DBD. Specifically, the predicted DBD mask from a generator is first exploited to re-generate two composite images by transporting the estimated clear and unclear areas from the source image to realistic full-clear and full-blurred images, respectively. To encourage these two composite images to be completely in-focus or out-of-focus, a global similarity discriminator is exploited to measure the similarity of each pair in a contrastive way, through which each two positive samples (two clear images or two blurred images) are enforced to be close while each two negative samples (a clear image and a blurred image) are inversely far. Since the global similarity discriminator only focuses on the blur-level of a whole image and there do exist some fail-detected pixels which only cover a small part of areas, a set of local similarity discriminators are further designed to measure the similarity of image patches in multiple scales. Thanks to this joint global and local strategy, as well as the contrastive similarity learning, the two composite images are more efficiently moved to be all-clear or all-blurred. Experimental results on real-world datasets substantiate the superiority of our proposed method both in quantification and visualization. The source code is released at: https://github.com/jerysaw/M2CS.

Early-Season Mapping of Winter Wheat and Garlic in Huaihe Basin Using Sentinel-1/2 and Landsat-7/8 Imagery

Early crop mapping is essential in predicting crop yield, assessing agricultural disasters, and responding to food price fluctuations. Winter wheat is a major food contributor in China. Existing early season maps of winter wheat strongly depend on the shape of the time series curve, which limits applicability on large scales. Besides, the effect of garlic on winter wheat mapping is often ignored. In this study, we determined how early we could identify winter crops (winter wheat and garlic) by examining time series of different lengths, and generated annual 30-m winter wheat and garlic map of the Huaihe basin using the Random Forest classifier, Sentinel-1/2, and Landsat-7/8 time-series imagery. The results showed that garlic could be identified at the end of November by using four composite images with overall accuracy (OA) of 0.88, followed by winter wheat recognizable at the end of January by using eight composite images with an OA of 0.91. The proposed framework can also be implemented in other regions and crops to generate early season distribution maps of different crops.

Deep Composite Face Image Attacks: Generation, Vulnerability and Detection

Face manipulation attacks have drawn the attention of biometric researchers because of their vulnerability to Face Recognition Systems (FRS). This paper proposes a novel scheme to generate Composite Face Image Attacks (CFIA) based on facial attributes using Generative Adversarial Networks (GANs). Given the face images corresponding to two unique data subjects, the proposed CFIA method will independently generate the segmented facial attributes, then blend them using transparent masks to generate the CFIA samples. We generate 526 unique CFIA combinations of facial attributes for each pair of contributory data subjects. Extensive experiments are carried out on our newly generated CFIA dataset consisting of 1000 unique identities with 2000 bona fide samples and 526000 CFIA samples, thus resulting in an overall 528000 face image samples. We present a sequence of experiments to benchmark the attack potential of CFIA samples using four different automatic FRS. We introduced a new metric named Generalized Morphing Attack Potential (G-MAP) to benchmark the vulnerability of generated attacks on FRS effectively. Additional experiments are performed on the representative subset of the CFIA dataset to benchmark both perceptual quality and human observer response. Finally, the CFIA detection performance is benchmarked using three different single image based face Morphing Attack Detection (MAD) algorithms. The source code of the proposed method together with CFIA dataset will be made publicly available: https://github.com/jagmohaniiit/LatentCompositionCode.

TNO or Comet? The Search for Activity and Characterization of Distant Object 418993 (2009 MS9)

The perihelion of the trans-Neptunian object (TNO) 2009 MS9 brings it close to the distance where some long-period comets are seen to become active. Knowing this, and the fact that this object appears to brighten in excess of its predicted nucleus brightness, suggests that 2009 MS9 has a delayed onset of activity brought on by the sublimation of a species more volatile than water. In this paper, we characterize 2009 MS9's physical properties and investigate potential outgassing through composite images, sublimation models, and measurements of spectral reflectivity. We find that deep composite images of the object at various epochs along its orbit show no evidence of dust yet place sensitive limits to the dust production. We estimate the nucleus radius to be 11.5 ± 3.5 km using thermal IR modeling from NEOWISE data and use this and data pre-perihelion to estimate a geometric albedo of 0.25. We compare a CO sublimation activity model to its post-perihelion heliocentric light curve and find that these data support an active fractional area of 5 × 10−6 assuming 2 μm–sized grains and other typical comet parameters. The spectral reflectivity of the surface materials obtained with the Gemini Observatory and CFHT at different epochs shows a reddening spectral slope. We compare the physical properties of 2009 MS9 to both TNO and comet populations and speculate that 2009 MS9's reddening may be due to the buildup of a dust mantle on the surface and could be an explanation of why TNOs exhibit a color bimodality.

Towards asteroid detection in microlensing surveys with deep learning

Asteroids are an indelible part of most astronomical surveys though only a few surveys are dedicated to their detection. Over the years, high cadence microlensing surveys have amassed several terabytes of data while scanning primarily the Galactic Bulge and Magellanic Clouds for microlensing events and thus provide a treasure trove of opportunities for scientific data mining. In particular, numerous asteroids have been observed by visual inspection of selected images. This paper presents novel deep learning-based solutions for the recovery and discovery of asteroids in the microlensing data gathered by the MOA project. Asteroid tracklets can be clearly seen by combining all the observations on a given night and these tracklets inform the structure of the dataset. Known asteroids were identified within these composite images and used for creating the labelled datasets required for supervised learning. Several custom CNN models were developed to identify images with asteroid tracklets. Model ensembling was then employed to reduce the variance in the predictions as well as to improve the generalization error, achieving a recall of 97.67%. Furthermore, the YOLOv4 object detector was trained to localize asteroid tracklets, achieving a mean Average Precision (mAP) of 90.97%. These trained networks will be applied to 16 years of MOA archival data to find both known and unknown asteroids that have been observed by the survey over the years. The methodologies developed can be adapted for use by other surveys for asteroid recovery and discovery.