Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add support for less or more frequent mAP calculation while training #4918

Merged
merged 12 commits into from
Sep 1, 2023

Conversation

mmaaz60
Copy link

@mmaaz60 mmaaz60 commented Feb 25, 2020

This pull request adds the flexibility to calculate mAP less or more frequently while training. The user can set the command line argument -mAP_epochs to decide after how many epochs mAP should be calculated. For example in order to calculate mAP for each 2 Epochs run the training with command darknet.exe detector train data/obj.data yolo-obj.cfg darknet53.conv.74 -map -mAP_epochs 2.

Note that by default the code calculates the mAP after each 4 epochs. So this feature particularly be useful in the scenarios where user is training on big one class dataset where the model may converge quickly, hence more frequent validation mAP calculation is required.

'-mAP_epochs' decide after how many epochs mAP should be calculated. Default value is 4 which means the mAP will be calculated after each 4 epochs.
Add instructions to use '-mAP_epochs' command line argument while training.
Monitor training, Add support to display remaining training time on training chart
Merge Changes from Base Repo (AlexeyAB/darknet)
Merge changes from alexeab/darknet
Merge Changes from AlexeAB/darknet
@Luux
Copy link

Luux commented Apr 11, 2020

I was on the verge of implementing this myself today. Guess I'm lucky, lol. :D

src/detector.c Outdated Show resolved Hide resolved
src/detector.c Outdated Show resolved Hide resolved
@cenit cenit merged commit 5a9970b into AlexeyAB:master Sep 1, 2023
19 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants