glenn-jocher
commented
3 years ago ๐ Hello! Thanks for asking about model anchors. YOLOv5 ๐ uses a new Ultralytics algorithm called autoanchor for anchor verification and generation before training starts.
Autoanchor will analyse your anchors against your dataset and training settings (like --img-size), and will adjust your anchors as necessary if it determines the original anchors are a poor fit, or if an anchor count was specified in your model.yaml rather than anchor values, i.e.
# Specify anchor count (per layer)
anchors: 3
# --OR-- Specify anchor values manually
anchors:
- [10,13, 16,30, 33,23] # P3/8
- [30,61, 62,45, 59,119] # P4/16
- [116,90, 156,198, 373,326] # P5/32When generating new anchors, autoanchor first applies a kmeans function against your dataset labels (scaled to your training --img-size), and then uses these kmeans centroids as initial conditions for a Genetic Evolution (GE) algorithm. The GE algorithm will evolve all anchors for 1000 generations under default settings, using the CIoU loss (same one used during training) combined with Best Possible Recall (BPR) as its fitness function.
No action is required on your part to use autoanchor. If you would like to force manual anchors for any reason, you can skip autoanchor with the --noautoanchor flag:
python train.py --noautoanchor
caihaunqai
github-actions[bot]
Hello, I want to use custom dataset and add to 15 anchors output. However, my precision and MAP falls when I use default P3 anchor size as P1 and P2's anchor sizes.
Is it the problem that my results worse? Do I have to calculate the custom dataset anchor size and how can I calculate them? Any demo is done provided? Thank you very much!