search

dyabel / detpro

Apache License 2.0
174 stars 28 forks source link

baiduyun link miss val part of lvis_clip_image_embedding #18

Closed ZhenghaoJi closed 2 years ago

ZhenghaoJi commented 2 years ago

lvis_clip_image_embedding.zip.xxx i cat them as lvis_clip_image_embedding.zip with 180g. after spending one day on solving my disk storage problem and unziping lvis_clip_image_embedding.zip, i get only tarin/train2017 without val/trainxx&val. I guess that when you compressing the files, val part was forgot.

ZhenghaoJi commented 2 years ago

and i found one roi_head.py inflating: data/lvis_clip_image_embedding/train2017/000000581929.pth
inflating: data/lvis_clip_image_embedding/train2017/standard_roi_head.py

dyabel commented 2 years ago

and i found one roi_head.py inflating: data/lvis_clip_image_embedding/train2017/000000581929.pth inflating: data/lvis_clip_image_embedding/train2017/standard_roi_head.py

Hi, there is no need to extract embedding for validation set here. The roi_head.py is a mistake, just ignore it.

ZhenghaoJi commented 2 years ago

so when cat 173file as one file lvis_clip_image_embedding.zip. put it under ./data in prepare.sh, what step should do next? run prompt_save_train_reuse.py and ./tools/dist_train.sh configs/lvis/prompt_save_val.py? I wanna reproduce vild* result without prompt.

CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/detpro_ens_20e.py 2 --work-dir workdirs/collect_data --cfg-options model.roi_head.load_feature=False total_epochs=1

compress the embeddings into a zip file

zip -r data/lvis_clip_image_embedding.zip data/lvis_clip_image_embedding/*

For training the DetPro, we need to save the embeddings of proposals embedded by the clip image encoder, correponding labels and ious with gt boxes.

CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/prompt_save_train.py 2 --work-dir workdirs/prompt_save_train CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/prompt_save_train_reuse.py 2 --work-dir workdirs/prompt_save_train CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/prompt_save_val.py 2 --work-dir workdirs/prompt_save_val python prompt/gather.py data/lvis_clip_image_proposal_embedding/train train_data.pth python prompt/gather.py data/lvis_clip_image_proposal_embedding/val val_data.pth

thanks a lot for your help! @dyabel

dyabel commented 2 years ago

so when cat 173file as one file lvis_clip_image_embedding.zip. put it under ./data in prepare.sh, what step should do next? run prompt_save_train_reuse.py and ./tools/dist_train.sh configs/lvis/prompt_save_val.py? I wanna reproduce vild* result without prompt.

CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/detpro_ens_20e.py 2 --work-dir workdirs/collect_data --cfg-options model.roi_head.load_feature=False total_epochs=1 #compress the embeddings into a zip file zip -r data/lvis_clip_image_embedding.zip data/lvis_clip_image_embedding/* #For training the DetPro, we need to save the embeddings of proposals embedded by the clip image encoder, correponding labels and ious with gt boxes. CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/prompt_save_train.py 2 --work-dir workdirs/prompt_save_train CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/prompt_save_train_reuse.py 2 --work-dir workdirs/prompt_save_train CUDA_VISIBLE_DEVICES=6,7 ./tools/dist_train.sh configs/lvis/prompt_save_val.py 2 --work-dir workdirs/prompt_save_val python prompt/gather.py data/lvis_clip_image_proposal_embedding/train train_data.pth python prompt/gather.py data/lvis_clip_image_proposal_embedding/val val_data.pth

thanks a lot for your help! @dyabel

For vild*, you only need to run the first two steps in prepare.sh to get lvis_clip_image_embedding.zip and use the class embedding I provided in the link, or extract it using manual prompts yourself.

Kegard commented 1 year ago

hello， I want to reproduce detpro but I cannot get the val/val2017, what should i do?