few-shot-object-detection
few-shot-object-detection copied to clipboard
How to run on seed 0?
Hello!
I was wondering what is the protocol to replicate your seed 0 results? I put the misc. txt files into a seed 0 folders and ran
python -m tools.run_experiments --num-gpus 1 --shots 1 --seeds 0 1 --split 1
but got a higher nAP50 (~45). For single GPU training I decreased my number of images to 2, the learning rate by 8, and increased the number of iters by 8 times.
Thank you!