{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":739471624,"defaultBranch":"main","name":"customer_identity_torch","ownerLogin":"MBTMBTMBT","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2024-01-05T16:41:39.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/59250708?v=4","public":true,"private":false,"isOrgOwned":false},"refInfo":{"name":"","listCacheKey":"v0:1713946682.0","currentOid":""},"activityList":{"items":[{"before":"1731b32fde7a5cdedcd4ed4ba617adeee57f1e12","after":"e92aef605e03c127ea8f47a8f951ecaa602a8f51","ref":"refs/heads/main","pushedAt":"2024-05-14T12:44:01.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Added DeepFashion2GeneralizedCategoriesAndAttributes class\n\nThis commit includes the implementation of a new class named DeepFashion2GeneralizedCategoriesAndAttributes. This class contains subsets of fashion categories, their respective attributes, and handles their threshold values. Besides, default thresholds and placeholders for specific thresholds were added.","shortMessageHtmlLink":"Added DeepFashion2GeneralizedCategoriesAndAttributes class"}},{"before":"81ab463c4158d4b5adb5491796be260b21d77f8e","after":"1731b32fde7a5cdedcd4ed4ba617adeee57f1e12","ref":"refs/heads/main","pushedAt":"2024-05-13T15:20:51.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Update bounding box and loss calculations in various scripts\n\nThis commit pertains to the improvement of bounding box adjustments and updates in the 'datasets.py' script. Moreover, the calculations for losses in the 'train.py' script have been optimized, outdating redundant lines. In addition, enhancements have been made to the display of category labels in the 'utils.py' script for better visualization, and adjustments have been made to the 'test_deepfashion2.py' script including the change of the model directory path and correct handling of prediction classes.","shortMessageHtmlLink":"Update bounding box and loss calculations in various scripts"}},{"before":"fc5e970866b375d4cdfb872fa4f62aecfb40a9e8","after":"81ab463c4158d4b5adb5491796be260b21d77f8e","ref":"refs/heads/main","pushedAt":"2024-05-10T13:43:58.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Refactor bounding box handling and loss calculation in training\n\nThe changes in this commit refactor the handling of bounding box tensors and the calculation of losses in the training process. By using the number of bounding box classes from the model and correctly sizing attribute masks, the commit ensures that loss calculations are more accurate and that irrelevant losses are correctly zeroed out. The Mean Average Precision (map) score calculation has also been updated to use properly sized prediction classes and attributes.","shortMessageHtmlLink":"Refactor bounding box handling and loss calculation in training"}},{"before":"16a56f378a317a26b2cdd7b7d45cbcba6e8916c8","after":"fc5e970866b375d4cdfb872fa4f62aecfb40a9e8","ref":"refs/heads/main","pushedAt":"2024-05-10T13:41:08.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Add training script for DeepFashion2 general labels\n\nAdded a detailed script for training DeepFashion2 general labels. This entails initializing models, datasets, and training metrics. Adjustments have also been made in datasets and training python files to fix minor issues and streamline the training process. The Dataset now returns 'general_bboxes' for bounding boxes rather than 'bboxes'. In the training script, bbox loss computation now correctly utilizes the number of bbox classes from the model.","shortMessageHtmlLink":"Add training script for DeepFashion2 general labels"}},{"before":"f4a891ccd82e1d6ba1055ca9c6dda74100655d41","after":"16a56f378a317a26b2cdd7b7d45cbcba6e8916c8","ref":"refs/heads/main","pushedAt":"2024-05-09T19:45:49.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Add general categories support to DeepFashion2Dataset\n\nThis commit extends the DeepFashion2Dataset with support for general categories. A new subclass, DeepFashion2DatasetGeneral, loads existing category data and additional general category data. This includes converting and processing category-specific and general masks. It also adjusts bounding boxes based on the additional general categories. The new subclass returns image tensors, masks, labels, and optional bounding boxes.","shortMessageHtmlLink":"Add general categories support to DeepFashion2Dataset"}},{"before":"161498e4cd6553b08bb53bfc16cdc63cc6f5b373","after":"f4a891ccd82e1d6ba1055ca9c6dda74100655d41","ref":"refs/heads/main","pushedAt":"2024-05-09T16:35:06.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Adjust true boxes selection and validate accuracy calculation\n\nThe selection of relevant true boxes in train.py now considers only the indices where true labels == 1, which improves accuracy of training. In train_deepfashion2.py, the validation accuracy is now correctly calculated as the average of average precision (mAP) and F1 score, and the image size is adjusted to 226x300.","shortMessageHtmlLink":"Adjust true boxes selection and validate accuracy calculation"}},{"before":"767b185c8985525131eae7251d24a14fe1aa43e9","after":"161498e4cd6553b08bb53bfc16cdc63cc6f5b373","ref":"refs/heads/main","pushedAt":"2024-05-09T14:42:05.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Update model parameters and logic in deepfashion2 scripts\n\nModified several parameters and logic in `test_deepfashion2.py` and `train_deepfashion2.py` such as model directory name, image size, scale range and label threshold for bounding boxes. Also switched the model from ResNet50 to ResNet18 in `models.py` and adjusted related configurations including decoder layers and Conv2d dimensions. These changes aim to improve the performance and accuracy of the model.","shortMessageHtmlLink":"Update model parameters and logic in deepfashion2 scripts"}},{"before":"cd5504f7dd13840c0b3bfffa1599322d707dbe65","after":"767b185c8985525131eae7251d24a14fe1aa43e9","ref":"refs/heads/main","pushedAt":"2024-05-03T17:01:23.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Update model structure and refine metrics calculation function\n\nThe commit updates the model to use resnet50 instead of resnet18, refactors the Decoder structure to match this change, and adds an extension to the bounding box generation part of the model. Additionally, the 'calculate_map' function has been improved to focus on boxes associated with label 1, and its use has been adapted across the training scripts. The commit also changes the model saving directory.","shortMessageHtmlLink":"Update model structure and refine metrics calculation function"}},{"before":"e136c6d290fe82eb672b1a134ca314434f91150b","after":"cd5504f7dd13840c0b3bfffa1599322d707dbe65","ref":"refs/heads/main","pushedAt":"2024-05-03T13:07:15.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Adjust loss calculation and optimizer learning rate\n\nIn this commit, the final bounding box loss value in the training script is scaled by a factor of 4. Additionally, commented out the calculation conversion of attributes and prediction classes to numpy arrays. In the DeepFashion2 training script, the learning rate of the Adam optimizer has been reduced from 5e-4 to 1e-4.","shortMessageHtmlLink":"Adjust loss calculation and optimizer learning rate"}},{"before":"4f7737aac91aaa3e430964518272ea92e0d6fe80","after":"e136c6d290fe82eb672b1a134ca314434f91150b","ref":"refs/heads/main","pushedAt":"2024-05-03T12:48:44.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Increase num_workers in data loaders\n\nThe number of workers used by the train_loader and val_loader in our deepfashion2 training script has been increased from 0 to 8 to speed up the data loading process. This change will enable the script to leverage the potential of multicore processors to load and preprocess data faster, potentially leading to shorter overall training times.","shortMessageHtmlLink":"Increase num_workers in data loaders"}},{"before":"48342c767d8c69fe69757e8125be3cce32df4a9e","after":"4f7737aac91aaa3e430964518272ea92e0d6fe80","ref":"refs/heads/main","pushedAt":"2024-05-03T12:47:48.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Refactor F1 score calculation in training and validation\n\nChanged the F1 score calculation in the training and validation process to use binary conversion before calculation. This improves the F1 metric accuracy as it works more effectively with binary labels for multi-label classification problems. Also, num_workers in DataLoader is set to 0 for both training and validation process in 'train_deepfashion2.py' to remove multi-processing.","shortMessageHtmlLink":"Refactor F1 score calculation in training and validation"}},{"before":"55587aea7bb848c529317557113f85972cde9f94","after":"48342c767d8c69fe69757e8125be3cce32df4a9e","ref":"refs/heads/main","pushedAt":"2024-05-03T11:31:12.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Adjust output size of penultimate layer in model\n\nThe output size of the penultimate layer of the model was previously set to 128. This has now been adjusted to 256 to match the size of the previous layers, ensuring consistency in the model output size.","shortMessageHtmlLink":"Adjust output size of penultimate layer in model"}},{"before":"bcf140bdce3d2cebfe8c9a58a65e122e15588845","after":"55587aea7bb848c529317557113f85972cde9f94","ref":"refs/heads/main","pushedAt":"2024-05-03T11:27:58.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Switch from ResNet50 to ResNet18 in segmentation models\n\nThe changes in this commit switch the base model from ResNet50 to ResNet18 in both SegmentPredictor and SegmentPredictorBbox classes. In addition, the corresponding decoder layers, classification head and bounding box generator configurations have been adjusted to match ResNet18 structure. These changes should make the models run faster with slightly less accuracy.","shortMessageHtmlLink":"Switch from ResNet50 to ResNet18 in segmentation models"}},{"before":"31cefcb5c18a070b0cbcd4786deafbdab6ec8bc7","after":"bcf140bdce3d2cebfe8c9a58a65e122e15588845","ref":"refs/heads/main","pushedAt":"2024-05-03T11:24:53.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Update f1_score average parameter and DataLoader num_workers\n\nThis commit changes the 'average' parameter in `f1_score()` from 'micro' to 'samples' in two instances within 'train.py'. It also increases the 'num_workers' value for `DataLoader()` from 0 to 8 in 'train_deepfashion2.py', potentially speeding up data loading within the training process.","shortMessageHtmlLink":"Update f1_score average parameter and DataLoader num_workers"}},{"before":"566b879e1610cc1eb1fd9f67caf040aeff04b008","after":"31cefcb5c18a070b0cbcd4786deafbdab6ec8bc7","ref":"refs/heads/main","pushedAt":"2024-05-03T11:20:44.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Updated deepfashion2 model and scripts to predict bounding boxes\n\nThe deepfashion2 model has been updated to predict bounding boxes in addition to masks and classes. This was achieved by changing the SegmentPredictor model to SegmentPredictorBbox in test_deepfashion2.py. Certain code blocks that were unnecessary or incorrect due to the addition of bounding boxes were corrected or removed. Changes were also made in train_deepfashion2.py script to correctly work with updated model and in models.py where modifications were made in bbox_generator's sequential model.","shortMessageHtmlLink":"Updated deepfashion2 model and scripts to predict bounding boxes"}},{"before":"c65e713ad64e9179b9b3c1150d7f9ccb2dce572f","after":"566b879e1610cc1eb1fd9f67caf040aeff04b008","ref":"refs/heads/main","pushedAt":"2024-05-02T13:29:23.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Add warning filters and comment out cudnn setting\n\nThe cudnn setting in 'train_deepfashion2' and 'train_celeba' files has been commented out. We also added a line to filter out warnings in both files to ensure smoother operation and cleaner console outputs.","shortMessageHtmlLink":"Add warning filters and comment out cudnn setting"}},{"before":"cbf69d54ac2bb15987c638356748318e90ce9da8","after":"c65e713ad64e9179b9b3c1150d7f9ccb2dce572f","ref":"refs/heads/main","pushedAt":"2024-05-01T23:58:17.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Uncomment code to calculate average f1 score\n\nThe previously commented line of code to calculate the average f1 score during validation has been uncommented. This will now allow the computation of the average f1 score along with other metrics during model validation.","shortMessageHtmlLink":"Uncomment code to calculate average f1 score"}},{"before":"efe264b6c4192a62cc8c3b6653f8fe8a4ccc273f","after":"cbf69d54ac2bb15987c638356748318e90ce9da8","ref":"refs/heads/main","pushedAt":"2024-05-01T23:57:48.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Updated calculation in training progress bar\n\nThe progress bar in train.py script has been updated to show the average losses and scores instead of instantaneous values. Additionally, box training loss has been added and average F1 score is now commented out.","shortMessageHtmlLink":"Updated calculation in training progress bar"}},{"before":"282b0191675c650de93fa9b46b1723d7aa9b392d","after":"efe264b6c4192a62cc8c3b6653f8fe8a4ccc273f","ref":"refs/heads/main","pushedAt":"2024-05-01T20:13:20.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Add backpropagation to training loss and update validation function\n\nThe commit includes the implementation of backpropagation in the calculation of the training loss. The validation function, val_DeepFashion2, has also been updated by removing the random scaling of images. Additionally, the validate function in train_deepfashion2.py has been replaced with val_DeepFashion2 to maintain consistency.","shortMessageHtmlLink":"Add backpropagation to training loss and update validation function"}},{"before":"21082a7f00cac8f189c55ddf789d9ecdca2bf71e","after":"282b0191675c650de93fa9b46b1723d7aa9b392d","ref":"refs/heads/main","pushedAt":"2024-04-30T23:20:40.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Updated \"iou\" and \"calculate_map\" functions to operate on batches\n\nThe \"iou\" and \"calculate_map\" functions have been updated to work on batches of boxes instead of individual ones. Additionally, these functions have been incorporated into the training and validation processes. A few other minor modifications have been made to ensure the code works correctly with these updates. For example, the \"calculate_map\" function now accepts predicted scores as an argument, and some tensor shape manipulations have been added for the correct computation.","shortMessageHtmlLink":"Updated \"iou\" and \"calculate_map\" functions to operate on batches"}},{"before":"eff2c208d91cf5fca2043844f1339133abe800b5","after":"21082a7f00cac8f189c55ddf789d9ecdca2bf71e","ref":"refs/heads/main","pushedAt":"2024-04-30T15:31:47.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Refactor training and validation processes to include bbox loss calculation\n\nThe code was refactored to include bounding box (bbox) loss calculation as part of the training and validation processes. This involves computing Mean Average Precision (mAP) and F1 scores as evaluation metrics. An Intersection over Union function was added to help calculate mAP, along with a function to calculate the mAP itself.","shortMessageHtmlLink":"Refactor training and validation processes to include bbox loss calcu…"}},{"before":"8c10825c4eb9d5aedf502669e9062229782cc950","after":"eff2c208d91cf5fca2043844f1339133abe800b5","ref":"refs/heads/main","pushedAt":"2024-04-30T14:01:08.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Add bbox prediction to SegmentPredictor\n\nImplemented a subclass of SegmentPredictor, SegmentPredictorBbox, which can predict bounding boxes in addition to masks and class labels. The bounding box predictions are generated using a series of linear layers, with output size equal to four times the number of bounding box classes.","shortMessageHtmlLink":"Add bbox prediction to SegmentPredictor"}},{"before":"3d7f5915ee9c5d077eaa51803709f48ae4d0287d","after":"8c10825c4eb9d5aedf502669e9062229782cc950","ref":"refs/heads/main","pushedAt":"2024-04-30T13:18:55.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Refactor code and optimise image segmentation\n\nThe image segmentation function has been reduced in size and moved to another file to streamline the code and improve its organisation. Furthermore, the number of data loads for the training dataset has been increased, which is expected to enhance the model's performance. Used Resnet50 instead of Resnet101 to improve speed.","shortMessageHtmlLink":"Refactor code and optimise image segmentation"}},{"before":"040df26d8dc2ec8458e697c03afb908c011e4d54","after":"3d7f5915ee9c5d077eaa51803709f48ae4d0287d","ref":"refs/heads/main","pushedAt":"2024-04-29T10:14:06.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Update ResNet model and adjust DataLoader parameters\n\nThe ResNet model in models.py has been switched from resnet152 to resnet101 to potentially enhance model performance. Furthermore, the arguments of DataLoader in train_deepfashion2.py have been adjusted, lowering the number of workers from 12 to 8 and altering the batch size for the validation loader, which could impact model training speed and memory usage. Batch normalization in the classifier has also been removed for model simplification.","shortMessageHtmlLink":"Update ResNet model and adjust DataLoader parameters"}},{"before":"b08fb0feb1c624b134d09d0f9a84e367402f9e23","after":"040df26d8dc2ec8458e697c03afb908c011e4d54","ref":"refs/heads/main","pushedAt":"2024-04-29T01:25:20.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Update resnet model and simplify train function call\n\nThe resnet model has been updated from resnet101 to resnet152 in models.py to possibly improve the prediction accuracy, and also the train function call has been simplified to be more readable in train_deepfashion2.py. This simplification does not alter any functionality but improves the readability and understandability of the code.","shortMessageHtmlLink":"Update resnet model and simplify train function call"}},{"before":"9682a729b0aa5c92247a6bef58243c6794f8d44f","after":"b08fb0feb1c624b134d09d0f9a84e367402f9e23","ref":"refs/heads/main","pushedAt":"2024-04-29T01:04:20.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Update training and validation process in deepfashion2\n\nChanges include updating the model from IntegratedModel to SegmentPredictor and removing the collate_fn from DataLoader. Also, updated the loss functions and metrics logged in TensorBoard. Corresponding training and validation functions were modified accordingly.","shortMessageHtmlLink":"Update training and validation process in deepfashion2"}},{"before":"9682ba9eb4615c41c40c335f6b0dedd66d2cb28f","after":"9682a729b0aa5c92247a6bef58243c6794f8d44f","ref":"refs/heads/main","pushedAt":"2024-04-29T01:03:59.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Optimize import order and handle model exceptions\n\nThe import order has been adjusted in the training script to follow the best practices. Also, the freeze and unfreeze segment model methods in the training loop are now wrapped in a try-except block to ignore possible exceptions and ensure smooth execution.","shortMessageHtmlLink":"Optimize import order and handle model exceptions"}},{"before":"69e888dd6a56baa5a9c316492bee7008c9ca8745","after":"9682ba9eb4615c41c40c335f6b0dedd66d2cb28f","ref":"refs/heads/main","pushedAt":"2024-04-29T01:03:47.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Add Faster R-CNN model in test_model.py\n\nThis commit introduces a Faster R-CNN model for object detection in test_model.py. The model uses ResNet50 as the backbone network, a Region Proposal Network (RPN), and a Fast R-CNN head. The code includes helper functions for generating anchor boxes and applying deltas, and a forward pass using a dummy image is demonstrated.","shortMessageHtmlLink":"Add Faster R-CNN model in test_model.py"}},{"before":"ba3d9ef3bbe63feddf95c083c2b49f8d4a3246c9","after":"69e888dd6a56baa5a9c316492bee7008c9ca8745","ref":"refs/heads/main","pushedAt":"2024-04-29T01:03:38.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Remove test_model.py in deepfashion2 directory\n\nThe test_model.py file in the deepfashion2 directory has been deleted. This file contained a model used for testing purposes, but was not essential for the final deployment of the program. It included a multitask ResNet50 model implementation, which is no longer needed in this directory.","shortMessageHtmlLink":"Remove test_model.py in deepfashion2 directory"}},{"before":"5b6fc9c428e976dfc8a8c44005729414089fe25e","after":"ba3d9ef3bbe63feddf95c083c2b49f8d4a3246c9","ref":"refs/heads/main","pushedAt":"2024-04-29T01:03:29.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"MBTMBTMBT","name":"MBT","path":"/MBTMBTMBT","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/59250708?s=80&v=4"},"commit":{"message":"Refactor Decoder and add SegmentPredictor class\n\nThe Decoder class in `models.py` has been simplified by streamlining the upsampling convolution operations. A major enhancement was made with the addition of the SegmentPredictor class, which leverages a pre-trained ResNet101 model for both semantic segmentation and classification tasks. It includes encoder, decoder, segmentation, and classification heads, as well as sigmoid activation for the output.","shortMessageHtmlLink":"Refactor Decoder and add SegmentPredictor class"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"Y3Vyc29yOnYyOpK7MjAyNC0wNS0xNFQxMjo0NDowMS4wMDAwMDBazwAAAARJdiML","startCursor":"Y3Vyc29yOnYyOpK7MjAyNC0wNS0xNFQxMjo0NDowMS4wMDAwMDBazwAAAARJdiML","endCursor":"Y3Vyc29yOnYyOpK7MjAyNC0wNC0yOVQwMTowMzoyOS4wMDAwMDBazwAAAAQ8gawx"}},"title":"Activity · MBTMBTMBT/customer_identity_torch"}