Why are the two results different in YOLOv5?

Question

I wanted to know the number of vehicles in the picture using yolov5 However, the result of the model was different from detect.py

0. img

1. model_result

# Model
model = torch.hub.load('ultralytics/yolov5', 'yolov5s')  # or yolov5m, yolov5l, yolov5x, custom

# Images
img = 'D:\code\YOLO\dataset\img\public02.png'  # or file, Path, PIL, OpenCV, numpy, list

# Inference
results = model(img)

# Results
results.print()  # or .show(), .save(), .crop(), .pandas(), etc.

result -> (no detections)

2. detect.py

from IPython.display import Image
import os

val_img_path = 'D:\code\YOLO\dataset\img\public02.png'
 
!python detect.py --img 416 --conf 0.25 --source "{val_img_path}"

result ->

I know that if I don't specify weight option in detect.py, the default yolo5s model is used. but, Result 1 differs from Result 2 using the same model.

Answer 1

It seems it is an image processing issue.

Indeed with your example, the model loaded from torch hub gives a different output than detect.py. Looking at the source code of detect.py I see that there is some good image pre-processing. From the model hub, I really don't know what's happening to the input. From the model hub, the resulting image is this:

With their pre-processing, this is basically the image that you are feeding into the model. Would not expect any detections from this to be honest.

But then I tried doing the pre-processing myself (noted as well in their tutorial)

import torch
import cv2

# Model
model = torch.hub.load('ultralytics/yolov5', 'yolov5s')  # or yolov5m, yolov5l, yolov5x, custom

# Image
imgPath = '/content/9X9FP.png'
img = cv2.imread(imgPath)[..., ::-1]  # Pre-processing OpenCV image (BGR to RGB)

# Inference
results = model(img)

# Results
results.save()

And it all works fine:

So for a quick and easy answer, I would just do the pre-processing my self, it's just a simple one line extra step. Good luck!