Support Objects365 pretrain and Adding the DINO++ model can achieve an accuracy of 63.4mAP at a model scale of 200M(Under the same scale, the accuracy is the best)
* add caltech, flower, mnist data source
* add det lvis data source
* add pose crowdPose data source
* add pose of OC Human data source
* add pose of mpii data source
* add Seg of voc data source
* add Seg of coco data source
* add Det of wider person datasource
* add Det of african wildlife datasource
* add Det of fruit datasource
* add Det of pet datasource
* add Det of artaxor and tiny person datasource
* add Det of wider face datasource
* add Det of crowd human datasource
* add Det of object365 datasource
* add Seg of coco stuff 10k and 164k datasource
Co-authored-by: Cathy0908 <30484308+Cathy0908@users.noreply.github.com>
1. Use original config as startup script. (For details, see refactor config parsing method #225)
2. Refactor the splicing rules of the check_base_cfg_path function in the EasyCV/easycv/utils/config_tools.py
3. Support three ways to pass class_list parameter.
4. Fix the bug that clsevalutor may make mistakes when evaluating top5.
5. Fix the bug that the distributed export cannot export the model.
6. Fix the bug that the load pretrained model key does not match.
7. support cls data source itag.
* add data_source imagenet
* modify data_source imagenet and add unittest
* modify data_source imagenet and modify unittest
* modify voc data_source and modify voc unittest and download Part
* modify coco data_source and modify coco unittest and add download Part , modify voc data_source
* add dataset metadata Format specification
* add pose download data , modify coco.py and modiy download file function ,add test coco download a part
* modify download file function
* modify download of cifar10 and cifar100
* modify dataset_name to target_dir
* create download_util and modify function
* modify function
* modify function
* add test case , modify
* modify something
* modify something
* modify something
* modify something
* modify something
* modify something
* add wget
* add wget
* modify
* add new modify
* add new modify
* modify something
* modify something and add something
* modify something and add something
* modify something and add something
* modify something and add something
* modify something and add something
* modify test case
* modify test case
* modify test case
* modify test case
* modify test case
1.Add a backbone: deitiii.
2.Add an optimizer: lamb.
3.Add a sampler: RASampler.
4.Add a lr update hook: CosineAnnealingWarmupByEpochLrUpdaterHook.
5.In easycv/models/classification/classification.py, I remove the default mixup_cfg to keep the classification.py clean.
when adapt for mmlab, the original modules will be changed, which will make argument error when using mmdet original config and interface
1. add interface for removing adaptation for mmlab
2. register copy of mmlab module into easycv
3. bump internal version to 0.6.3.1
Link: https://code.alibaba-inc.com/pai-vision/EasyCV/codereview/10059050
* complete yolox introduction
* fix env problem with nni2.6.1
* fix yolox-edge problem
* add docker turtorial
* add blade test
* support old export yolox model
Co-authored-by: wuziheng <zhoulou.wzh@alibaba-inc.com>
