[Learning a unified label space]: FileNotFoundError: [Errno 2] No such file or directory for "json.load(open(unified_label_file, 'r'))"

[jichengyuan]

Hi,

thank you for this awesome idea on this nice multi-dataset-object-detectror.

I've trained a "Partitioned detector" on coco, oid and obejcts365-V2. (for V2, because MEGVII have updated their dataset, we could not download v1 anymore.)

When I tried the second step: Learning a unified label space:
I followed the tutorial and run:

python projects/UniDet/train_net.py \
--config-file projects/UniDet/Partitioned_COI_R50_2x.yaml \
--num-gpus 1 \
--eval-only MULTI_DATASET.UNIFIED_EVAL True

It raised a "FileNotFoundError". When I checked config file, I found that the key "UNIFIED_LABEL_FILE" in the config file has been set as ' ' (None), as shown in images below.

Could you give me some hints about how could I generate this this file or label space?

many thanks!

FileNotFoundError:

config.yaml:

[jichengyuan]

I also tried with the command below for “Learning a unified label space”:

python projects/UniDet/train_net.py \
--config-file projects/UniDet/Partitioned_COI_R50_2x.yaml \
--num-gpus 1 \
--eval-only \
MULTI_DATASET.UNIFIED_EVAL True \
MODEL.WEIGHTS output/UniDet/Partitioned_COI_R50_2x/model_final.pth \
MULTI_DATASET.UNIFIED_LABEL_FILE datasets/label_spaces/manual.json

Then I got very low mAP:

but when I validated the model as partitioned detectors, it works fine:

python projects/UniDet/train_net.py \
--config projects/UniDet/configs/Partitioned_COI_R50_2x.yaml \
--eval-only \
--num-gpus 1 \
MODEL.WEIGHTS output/UniDet/Partitioned_COI_R50_2x/model_final.pth

[HXZhong1997]

Same question when I tried to train on my two custom datasets. Didn't have a clue yet 😢

[jichengyuan]

I have checked this file "projects\UniDet\unidet\evaluation\multi_dataset_evaluator.py"

For the "_unified_results", the label_space is a specific one (such as coco) rather than a unified label space.

That means, if we execute an evaluation on coco dataset, the "_unified_results" from the unified detector is same as the the "coco_results" from the partitioned detector.