[X] I have searched the YOLOv5 issues and discussions and found no similar questions.
Question
Hi @glenn-jocher ,
I trained a custom data using yolov5m pre-trained weight to detect weather a person is pedestrian or not. My data has two classes. I use minibatch size = 16 and epochs=100. These are the hyperparameter I used:
lr0: 0.01
lrf: 0.01
momentum: 0.937
weight_decay: 0.0005
warmup_epochs: 3.0
warmup_momentum: 0.8
warmup_bias_lr: 0.1
box: 0.05
cls: 0.5
cls_pw: 1.0
obj: 1.0
obj_pw: 1.0
iou_t: 0.2
anchor_t: 4.0
fl_gamma: 0.0
hsv_h: 0.015
hsv_s: 0.7
hsv_v: 0.4
degrees: 0.0
translate: 0.1
scale: 0.5
shear: 0.0
perspective: 0.0
flipud: 0.0
fliplr: 0.5
mosaic: 1.0
mixup: 0.0
copy_paste: 0.0
I have 2550 train image and 451 validation image. I also added 2% background image with no labels as per previous issues suggestions.
I got only 0.508 mAP @0.5 and very low recall value. From result, I can also see the val/obj loss and val/class loss is increasing.
Also from confusion matrix, I got 0% correctly predicted background which is very confusing me.
Could you give any suggestion or direction to increase mAP value,please?
Search before asking
Question
Hi @glenn-jocher , I trained a custom data using yolov5m pre-trained weight to detect weather a person is pedestrian or not. My data has two classes. I use minibatch size = 16 and epochs=100. These are the hyperparameter I used: lr0: 0.01 lrf: 0.01 momentum: 0.937 weight_decay: 0.0005 warmup_epochs: 3.0 warmup_momentum: 0.8 warmup_bias_lr: 0.1 box: 0.05 cls: 0.5 cls_pw: 1.0 obj: 1.0 obj_pw: 1.0 iou_t: 0.2 anchor_t: 4.0 fl_gamma: 0.0 hsv_h: 0.015 hsv_s: 0.7 hsv_v: 0.4 degrees: 0.0 translate: 0.1 scale: 0.5 shear: 0.0 perspective: 0.0 flipud: 0.0 fliplr: 0.5 mosaic: 1.0 mixup: 0.0 copy_paste: 0.0 I have 2550 train image and 451 validation image. I also added 2% background image with no labels as per previous issues suggestions. I got only 0.508 mAP @0.5 and very low recall value. From result, I can also see the val/obj loss and val/class loss is increasing. Also from confusion matrix, I got 0% correctly predicted background which is very confusing me. Could you give any suggestion or direction to increase mAP value,please?
Additional
No response