TASK1:
- Git to get our repo
- Change the darknet folder's name into darknet_my
- Download deploy.zip in the folder
- Unzip the deploy.zip
- run "python3 gen_voc_label.py"
- get the darknet: git clone https://github.com/pjreddie/darknet
- cd darkent
- replace the Makefile in the darknet by the Makefile under the darknet_my
- make
- put the file under darknet_my/cfg/ into darknet/cfg/
- put the file under darknet_my/data/ into darknet/data/
- replace the detector.c under darknet/example/ by the detector.c under darkent_my/example/
- run "wget https://pjreddie.com/media/files/darknet53.conv.74"
- run "./darknet detector train cfg/my.data cfg/yolov3-my.cfg darknet53.conv.74"
- You will see the weight in ./backup/ after every 100 iteration
- run "./darkent detector valid cfg/my.data cfg/yolov3-mytest.cfg ./backup/yolov3-my_xxxx.weights > result.txt"
- You will see the prediction result for the test data
- Change the format according and can be submited (18. if you want to change the resolution for training, also should run python gen_anchor.py to generate the anchor and replace it in the .cfg)
TASK2:
- uncommand line 493, line 497 in detector.c under example, and command line 494, line 498 in detector.c
- cd darknet
- make
- run "./darkent detector valid cfg/my.data cfg/yolov3-mytest.cfg ./backup/yolov3-my_xxxx.weights > result_new.txt"
- You will get the result_new.txt for 2d object box position and size
- cd ..
- run "python3 task2_code.py"
- You will get the TASK2.txt for submission