Yolov5-Lite: Minimal YoloV5 Implementation
Project description
Yolov5-Lite: Minimal YoloV5 Implementation
Overview
It has been simplified by editing detect.py in the yolov5 repository.
Installation
git clone https://github.com/kadirnar/yolov5-lite
cd yolov5-lite
pip install -r requirements.txt
Yolov5-Lite Prediction:
It is the edited version of the codes in the detect file.
class Yolov5:
def __init__(self, weights, device, data):
self.weights = weights
self.device = device
def load_model(self, weights, device, data):
self.device = select_device(device)
self.model = DetectMultiBackend(weights, device=self.device, data=data)
def preprocces_img(self, img, imgsz):
self.npy_im = numpy_img(img, imgsz)
self.tensor_im = file_to_torch(self.npy_im, self.device)
def detect(self):
# Inference
pred = self.model(self.tensor_im) # shape: torch.Size([1, 3, 640, 480])
# NMS
pred = non_max_suppression(pred, conf_thres=0.25, iou_thres=0.45, max_det=1000)
for det in pred:
det[:, :4] = scale_coords(self.tensor_im.shape[2:], det[:, :4], self.npy_im.shape).round()
self.det = det
def show_img(self, view_img=True):
# Write results
for *xyxy, conf, cls in reversed(self.det):
annotator = Annotator(self.npy_im, line_width=3, example=str(self.model.names))
logging.info("\t+ Label: %s, Conf: %.5f" % (self.model.names[int(cls)], conf.item()))
if view_img: # Add bbox to image
label = "%s %.2f" % (self.model.names[int(cls)], conf)
annotator.box_label(xyxy, label, color=colors(int(cls), True))
# Stream results
im0 = annotator.result()
if view_img:
cv2.imshow("frame", im0)
cv2.waitKey(0)
cv2.destroyAllWindows()
Yolov5-Lite Run Code:
You can take the detect.py file as an example to load and visualize your yolov5 models.
weights = "yolov5s.pt"
img = "data/images/bus.jpg"
data = "data/coco128.yaml"
device = "cpu"
imgsz = 640
view_img = True
model = Yolov5(weights, device, data)
model.load_model(weights, device, data)
model.preprocces_img(img, imgsz)
model.detect()
model.show_img(view_img)
Reference:
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
yolite-0.0.3.tar.gz
(3.2 kB
view details)
File details
Details for the file yolite-0.0.3.tar.gz
.
File metadata
- Download URL: yolite-0.0.3.tar.gz
- Upload date:
- Size: 3.2 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.0 CPython/3.10.4
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 04066f10b93260b79b5929e0cb5e19b49d1136c78e25611e34d94cd92a8327c0 |
|
MD5 | 9738f86f260af7e04c323fd5ee05685d |
|
BLAKE2b-256 | d0b99a6e890d85ffce5527c2d449ba148c1a048cee25dc6bbaf9fa53f78e9ec0 |