mAP Score Low on Custom Data

eriklindernoren / PyTorch-YOLOv3

Minimal PyTorch implementation of YOLOv3

GNU General Public License v3.0

7.29k stars 2.64k forks source link

mAP Score Low on Custom Data #462

Open JamMochi opened 4 years ago

JamMochi commented 4 years ago

I've been trying to train yolo to detect infrared images. The data set i'm using is the FLIR Thermal Dataset. After training the model and running it through the test.py, I get a mAP score of like 0.002 or something around those ranges. I can't really get a grasp on what I'm doing wrong. Could anyone give me some insight especially if you train on custom data or the FLIR Thermal Dataset.

Thank You

kashishoberoi commented 4 years ago

I am getting the same issue when I am training on Indian Driving Dataset! Did you resolve it?

karthikiitm87 commented 6 months ago

Hi @kashishoberoi, did you try resolving the issue you were getting for IDD?