What does the output model mean after each script

[Parsifal133]

Hi, thanks for your ama~~~zing job!

I'd like to reproduce the 10+10 experiment according to Table 2 in your paper, which means i am running the scripts as follow:

(Base 10) python tools/train_net.py --num-gpus 1 --config-file ./configs/PascalVOC-Detection/iOD/base_10.yaml SOLVER.IMS_PER_BATCH 1 SOLVER.BASE_LR 0.005

(10 + 10) python tools/train_net.py --num-gpus 1 --config-file ./configs/PascalVOC-Detection/iOD/10_p_10.yaml SOLVER.IMS_PER_BATCH 1 SOLVER.BASE_LR 0.005

(10 + 10 _ ft) python tools/train_net.py --num-gpus 1 --config-file ./configs/PascalVOC-Detection/iOD/ft_10_p_10.yaml SOLVER.IMS_PER_BATCH 1 SOLVER.BASE_LR 0.005

(because of limited devices, I set --num-gpus and bs to 1. ) According to my understanding, the Base 10 script conducts the First 10 class training, refer to the 3-th line of table. The trained model can only detect the first 10 class.

However, the 10+10 script output a model that can only detect the last 10 class object, is it not the final model after continual learning? And, what does the 10+10_ft do, though i am running this script.

Look forward to your reply！

[Parsifal133]

Does the base 10 script train a model refer to the First 10 in Table 2?

[Parsifal133]

In one word, I don't know clearly what does the above three scripts mean or how can I reproduce the results in Table 2. Thanks for any advice!