achen353 / Taiwanese-Traffic-Object-Detection

Training and fine-tuning YOLOv4 Tiny on custom object detection dataset for Taiwanese traffic
Other
10 stars 3 forks source link

Please provide obj.data and obj.names #4

Open sriks6711 opened 8 months ago

sriks6711 commented 8 months ago

To make the command [1] run successfully, we also need obj.data and obj.names.

[1] $ ./darknet detector demo data/obj.data cfg/yolov4-tiny-obj.cfg yolov4-tiny-obj_best.weights

sriks6711 commented 8 months ago

To make the command [1] run successfully, we also need obj.data and obj.names.

[1] $ ./darknet detector demo data/obj.data cfg/yolov4-tiny-obj.cfg yolov4-tiny-obj_best.weights

@jaydubai While only @achen353 can provide them, here is a placeholder working somewhat -

1) obj.data (a simple text file with the following contents) classes=4 names=data/obj.names

2) obj.names (a simple text file with the following contents which contains the 4 classes) vehicle scooter pedestrian bicycle

jaydubal commented 8 months ago

Thanks, still not working on my end. Even when using the obj.names you suggested. I guess the main problem is the missing .cfg Which .cfg are you using?

xinthemayanksachan commented 2 months ago

Below are the changes that we have made for successful execution.

Few Modifications we have Made:

A. In File cfg/yolov4-tiny-custom.cfg:
    1. Replace filters=255 to filters=27 at line 212 and 361
    2. Replace classes=80 to classes=4 at line 218 and 267

B. In File cfg/coco.data:
    1. Replace classes= 80 to classes= 4 at line 1

C. In File data/coco.names:
    Replace the file contents as below:
        vehicle
        scooter
        pedestrian
        bicycle

After above modifications are made, ececute the below command: ./darknet detector demo cfg/coco.data cfg/yolov4-tiny-custom.cfg backup/yolov4-tiny-obj_best.weights sample_input video eg: ABC.mp4

jaydubal commented 2 months ago

Yes it is working now... the line 361 is actually 263 in .cfg file