https://github.com/cccaaannn/yolo_predictor
Predict images with tensorflow converted darknet yolov4 model in a single line.
https://github.com/cccaaannn/yolo_predictor
predict-images tensorflow yolo-predictor yolov4
Last synced: 2 months ago
JSON representation
Predict images with tensorflow converted darknet yolov4 model in a single line.
- Host: GitHub
- URL: https://github.com/cccaaannn/yolo_predictor
- Owner: cccaaannn
- License: mit
- Created: 2021-01-07T02:15:26.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2021-01-16T05:13:49.000Z (over 5 years ago)
- Last Synced: 2025-06-04T12:16:51.071Z (about 1 year ago)
- Topics: predict-images, tensorflow, yolo-predictor, yolov4
- Language: Python
- Homepage:
- Size: 1.9 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Yolo predictor
### Predict images with tensorflow converted darknet yolov4 model in a single line.
---
  [](https://github.com/cccaaannn/yolo_predictor/blob/master/LICENSE)
## Before starting
- Yolo predictor predicts single or multiple images with [darknet yolo](https://github.com/AlexeyAB/darknet) model
- Model has to be converted to tensorflow, you can use this repo for converting the model [github.com/hunglc007/tensorflow-yolov4-tflite](https://github.com/hunglc007/tensorflow-yolov4-tflite)
- Tested with tensorflow 2.4.0
## Usage
### Predict a single image
```shell
python predict.py -m model_files/yolov4_coco -n model_files/coco.names -i test_images/dog.jpg --show
```
### Predict a directory of images
```shell
python predict.py -m model_files/yolov4_coco -n model_files/coco.names -d test_images --show
```
### Save predicted images
```shell
python predict.py -m model_files/yolov4_coco -n model_files/coco.names -d test_images --save_folder test_results
```
### All arguments
```
Model arguments:
-m, --model_path Tensorflow converted saved model FOLDER path
-n, --names_path Class (names) file path
Image arguments:
-d, --dir_path Directory of the images
-i, --image_path Path of the image
Draw arguments:
-s, --show Draw and show images with bounding boxes
--save_folder Save folder for drawn images with bounding boxes
--resize Resize images with given dim, Ex: 1280 720
--suffix Saved file suffix. Ex: '_predicted' dog.jpg -> dog_predicted.jpg
```
## Example results
### For more flexibility you can use yolo_predictor class
```python
from yolo_predictor import yolo_predictor
from yolo_drawer import yolo_drawer
model_path = "model_files/yolov4_coco.weights"
names_path = "model_files/coco.names"
image_path = "test_images/dog.jpg"
# init model
predictor = yolo_predictor(model_path, names_path)
# predict
predictions = predictor.predict(image_path)
print(predictions)
# draw image
drawer = yolo_drawer()
image, save_path = drawer.draw(predictions, image_path, show=False, resize=False, save_folder_path="test_results")
import cv2
cv2.imshow("Predicted " + save_path, image)
cv2.waitKey()
```
### Output
```
[(class_name, class_index, confidence, (x, y, w, h)), (class_name, class_index, confidence, (x, y, w, h)), ...]
```
```shell
[('bicycle', 1, 0.9867098, (0.4529685378074646, 0.48244842886924744, 0.573084, 0.5117002)), ('dog', 16, 0.98514426, (0.28938552737236023, 0.6685629785060883, 0.23469335, 0.5305287)), ('truck', 7, 0.92009175, (0.754764.7547647655010223, 0.2147115096449852, 0.296138, 0.16953142))]
```