yolov5 - unable to do inference on custom model. Works fine on cli command line. But would like to run interactive but image not detected.

[jjwallaby]

Hi,
yolov5 - unable to do inference on custom model. Works fine on cli command line. But would like to run interactive.
This works :
from yolov5 subforlder.
python detect.py --weights best.pt --source image1.jpg --data data.yaml --img 320 --conf 0.7 ---- WORKS!!!
detect.py will use ./data/coco128.yaml by detault.

Output from detect.py:
(tensorflow39) PS D:\tensorflow39\license_driver_v2_f\yolov5> python detect.py --weights best.pt --source image1.jpg --data data.yaml --img 320 --conf 0.7
detect: weights=['best.pt'], source=image1.jpg, data=data.yaml, imgsz=[320, 320], conf_thres=0.7, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_csv=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs\detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False, vid_stride=1
YOLOv5 v7.0-227-ge4df1ec Python-3.9.18 torch-2.1.0+cu118 CUDA:0 (Quadro P620, 4096MiB)

Fusing layers...
custom_YOLOv5s summary: 157 layers, 7012822 parameters, 0 gradients, 15.8 GFLOPs
image 1/1 D:\tensorflow39\license_driver_v2_f\yolov5\image1.jpg: 320x320 1 driverlicense, 18.5ms
Speed: 0.0ms pre-process, 18.5ms inference, 2.2ms NMS per image at shape (1, 3, 320, 320)
Results saved to runs\detect\exp3

model = torch.hub.load('./yolov5', 'custom', path='./yolov5/best.pt', source='local', force_reload=True, autoshape=True)

What does this custom mean. When i trained the model I used --cfg custom_yolov5s.yaml

Result:

YOLOv5 v7.0-227-ge4df1ec Python-3.9.18 torch-2.1.0+cu118 CUDA:0 (Quadro P620, 4096MiB)

Fusing layers...
custom_YOLOv5s summary: 157 layers, 7012822 parameters, 0 gradients, 15.8 GFLOPs
Adding AutoShape...
from PIL import Image
img = Image.open('./yolov5/image1.jpg')
new_image = img.resize((320, 320))
results = model(new_image)
print(f'prediction: { results.pred}')
result:prediction: [tensor([], device='cuda:0', size=(0, 6))]

image 1/1: 320x320 (no detections)
Speed: 18.5ms pre-process, 132.3ms inference, 0.0ms NMS per image at shape (1, 3, 640, 640)