Toolkit
Toolkit for Deep Learning.
.
| - model/
|- vgg.py
|- resnet.py
|- densenet.py
|- fcn8s.py
|- unet.py
|- pspnet.py
|- deeplabv3+ (deeplabv3+.py, resnet_deeplab.py, aspp.py, decoder.py)
|- DualGCN.py
|- JL_DCF.py
|- swin transformer (swin_transformer_seg.py, mlp_decoder.py)
|- source-code/
|- bn_details.py
|- bn_run.py
|- CrossEntropyLoss.py
|- regularization.py
|- SoftMax.py
|- utils/
|- data
|- data_config.py
|- divide_data.py
|- img_ops.py
|- imgs2video.py
|- nms
|- nms_cpu.py
|- ...
|- bbox_iou.py
|- bbox_iou_python.py
|- BN_torch.py
|- count_norm.py
|- dataset.py
|- loss.py
|- seg_transform.py
|- show_img.py
|- README.md
|- test.py
|- test_net.py
|- train.py
model
model/vgg.py
: ICLR(2015) paper.
model/resnet.py
: CVPR(2016) paper.
model/densenet.py
: CVPR(2017) paper.
model/fcn8s.py
: CVPR(2015) paper.
model/unet.py
: MICCAI(2015) paper.
model/pspnet.py
: CVPR(2017) paper.
model/deeplabv3+
: ECCV(2018) paper.
model/DualGCN.py
: BMVC(2019) paper.
model/JL_DCF.py
: CVPR(2020) paper.
model/swin transformer
: arXiv(2021) paper.
source-code
source-code/bn_details.py
: implementation of BN(BatchNormalization) and analysis of its details.
source-code/bn_run.py
: implementation of BN(BatchNormalization) and simulation of running.
source-code/CrossEntropyLoss.py
: implementation of custom CrossEntropyLoss and BCELoss.
source-code/regularization.py
: implementation of L1/L2 normalization, L1/L2 regularization and Dropout.
source-code/SoftMax.py
: implementation of SoftMax function in various version.
utils
utils/data/count_nrom.py
: count the mean and the standard deviation from datasets.
utils/data/data_config.py
: config (hyper-)parameters in main.py.
utils/data/dataset.py
: implementation of dataset in segmentation.
utils/data/divide_data.py
: divide origin data into train and valid set.
utils/data/img_ops.py
: resize, hist equalize and blur images using opencv.
utils/data/imgs2video.py
: convert images to video and extract images from video.
utils/data/seg_transform.py
: implementation of transform module in segmentation.
utils/nms/nms_cpu.py
: remove useless bounding-box by nms(Non-maximum suppression).
utils/bbox_iou.py
: calculate iou(Intersection-of-Union) between two bounding-box.
utils/bbox_iou_python.py
: calculate iou in python version.
utils/logging_util.py
: implementation of logging module with formatting.
utils/loss.py
: implementation of loss function, including FocalLoss, BinaryDiceLoss.
utils/show_img.py
: visualize image in both plt(matplotlib.pyplot) and cv2(opencv).
(Updating...)