Spaces:
Sleeping
Sleeping
| #!/usr/bin/env python | |
| from __future__ import annotations | |
| import pathlib | |
| import cv2 | |
| import gradio as gr | |
| import huggingface_hub | |
| import insightface | |
| import numpy as np | |
| import onnxruntime as ort | |
| from PIL import Image | |
| import io | |
| TITLE = "insightface Person Detection" | |
| DESCRIPTION = "https://github.com/deepinsight/insightface/tree/master/examples/person_detection" | |
| def load_model(): | |
| path = huggingface_hub.hf_hub_download("public-data/insightface", "models/scrfd_person_2.5g.onnx") | |
| options = ort.SessionOptions() | |
| options.intra_op_num_threads = 8 | |
| options.inter_op_num_threads = 8 | |
| session = ort.InferenceSession( | |
| path, sess_options=options, providers=["CPUExecutionProvider", "CUDAExecutionProvider"] | |
| ) | |
| model = insightface.model_zoo.retinaface.RetinaFace(model_file=path, session=session) | |
| return model | |
| def detect_person( | |
| img: np.ndarray, detector: insightface.model_zoo.retinaface.RetinaFace | |
| ) -> tuple[np.ndarray, np.ndarray]: | |
| bboxes, kpss = detector.detect(img) | |
| bboxes = np.round(bboxes[:, :4]).astype(int) | |
| kpss = np.round(kpss).astype(int) | |
| kpss[:, :, 0] = np.clip(kpss[:, :, 0], 0, img.shape[1]) | |
| kpss[:, :, 1] = np.clip(kpss[:, :, 1], 0, img.shape[0]) | |
| vbboxes = bboxes.copy() | |
| vbboxes[:, 0] = kpss[:, 0, 0] | |
| vbboxes[:, 1] = kpss[:, 0, 1] | |
| vbboxes[:, 2] = kpss[:, 4, 0] | |
| vbboxes[:, 3] = kpss[:, 4, 1] | |
| return bboxes, vbboxes | |
| def visualize(image: np.ndarray, bboxes: np.ndarray, vbboxes: np.ndarray) -> np.ndarray: | |
| res = image.copy() | |
| for i in range(bboxes.shape[0]): | |
| bbox = bboxes[i] | |
| vbbox = vbboxes[i] | |
| x1, y1, x2, y2 = bbox | |
| vx1, vy1, vx2, vy2 = vbbox | |
| cv2.rectangle(res, (x1, y1), (x2, y2), (0, 255, 0), 1) | |
| alpha = 0.8 | |
| color = (255, 0, 0) | |
| for c in range(3): | |
| res[vy1:vy2, vx1:vx2, c] = res[vy1:vy2, vx1:vx2, c] * alpha + color[c] * (1.0 - alpha) | |
| cv2.circle(res, (vx1, vy1), 1, color, 2) | |
| cv2.circle(res, (vx1, vy2), 1, color, 2) | |
| cv2.circle(res, (vx2, vy1), 1, color, 2) | |
| cv2.circle(res, (vx2, vy2), 1, color, 2) | |
| return res | |
| def extract_persons(image: np.ndarray, bboxes: np.ndarray) -> list[np.ndarray]: | |
| person_images = [] | |
| for bbox in bboxes: | |
| x1, y1, x2, y2 = bbox | |
| person_image = image[y1:y2, x1:x2] # Crop the detected person | |
| person_images.append(person_image) | |
| return person_images | |
| def convert_to_pil_image(image: np.ndarray) -> Image.Image: | |
| """Convert a NumPy image array to a PIL Image.""" | |
| return Image.fromarray(image) | |
| def convert_to_png(image: np.ndarray) -> bytes: | |
| """Convert a NumPy image array to PNG bytes.""" | |
| pil_image = convert_to_pil_image(image) | |
| buffer = io.BytesIO() | |
| pil_image.save(buffer, format="PNG") | |
| return buffer.getvalue() | |
| detector = load_model() | |
| detector.prepare(-1, nms_thresh=0.5, input_size=(640, 640)) | |
| def detect(image: np.ndarray) -> tuple[bytes, list[bytes]]: | |
| image = image[:, :, ::-1] # RGB -> BGR | |
| bboxes, vbboxes = detect_person(image, detector) | |
| res = visualize(image, bboxes, vbboxes) | |
| person_images = extract_persons(image, bboxes) | |
| # Convert the images to PNG bytes | |
| result_image_bytes = convert_to_png(res[:, :, ::-1]) # BGR -> RGB | |
| person_images_bytes = [convert_to_png(person_img[:, :, ::-1]) for person_img in person_images] # BGR -> RGB | |
| return result_image_bytes, person_images_bytes | |
| examples = sorted(pathlib.Path("images").glob("*.jpg")) | |
| demo = gr.Interface( | |
| fn=detect, | |
| inputs=gr.Image(label="Input", type="numpy"), | |
| outputs=[gr.Image(label="Processed Image", type="bytes"), gr.Gallery(label="Detected Persons", type="bytes")], | |
| examples=examples, | |
| examples_per_page=30, | |
| title=TITLE, | |
| description=DESCRIPTION, | |
| ) | |
| if __name__ == "__main__": | |
| demo.queue(max_size=10).launch() | |