Upload 2 files

app.py

@@ -0,0 +1,57 @@
+import gradio as gr
+import mediapipe as mp
+import cv2
+import numpy as np
+from PIL import Image
+from transformers import AutoTokenizer, AutoProcessor, AutoModelForVision2Seq
+
+model_id = "Qwen/Qwen2-VL-7B"
+
+tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
+model = AutoModelForVision2Seq.from_pretrained(
+    model_id, trust_remote_code=True, device_map="auto"
+).eval()
+
+mp_pose = mp.solutions.pose
+
+def analyze_posture_by_keypoints(landmarks):
+    left_shoulder = landmarks.landmark[11]
+    right_shoulder = landmarks.landmark[12]
+    left_ear = landmarks.landmark[7]
+    right_ear = landmarks.landmark[8]
+
+    shoulder_x = (left_shoulder.x + right_shoulder.x) / 2
+    ear_x = (left_ear.x + right_ear.x) / 2
+    delta = ear_x - shoulder_x
+    if abs(delta) > 0.06:
+        return "该用户存在驼背或低头倾向，头部明显前倾。"
+    else:
+        return "该用户坐姿较为端正，头部与肩部对齐。"
+
+def process(image: Image):
+    np_image = np.array(image)
+    with mp_pose.Pose(static_image_mode=True) as pose:
+        results = pose.process(cv2.cvtColor(np_image, cv2.COLOR_RGB2BGR))
+        if not results.pose_landmarks:
+            return "❗ 无法检测到人体，请上传包含上半身的清晰坐姿照片。"
+
+        posture_analysis = analyze_posture_by_keypoints(results.pose_landmarks)
+        prompt = f"请根据以下坐姿描述生成中英文提醒：\n{posture_analysis}"
+        inputs = processor(images=image, text=prompt, return_tensors="pt").to(model.device)
+        outputs = model.generate(**inputs, max_new_tokens=512)
+        result = tokenizer.decode(outputs[0], skip_special_tokens=True)
+        return result
+
+demo = gr.Interface(
+    fn=process,
+    inputs=gr.Image(type="pil", label="上传你的坐姿照片"),
+    outputs=gr.Textbox(label="中英文坐姿分析结果"),
+    title="🪑 Qwen2-VL 坐姿识别融合助手",
+    description="Mediapipe + Qwen2-VL 模型融合，上传图像判断是否驼背，并生成中英文提醒。",
+    theme="soft",
+    allow_flagging="never"
+)
+
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,10 @@
+transformers==4.37.2
+torch
+torchvision
+gradio==4.16.0
+mediapipe
+opencv-python
+Pillow
+tiktoken
+einops
+transformers_stream_generator