Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from flask import Flask, request, jsonify
|
2 |
+
import torch
|
3 |
+
from transformers import AutoModelForSequenceClassification, AutoTokenizer
|
4 |
+
|
5 |
+
# Load Meta Sapiens Pose model
|
6 |
+
sapiens_model = torch.jit.load('/models/sapiens_pose/model.pt')
|
7 |
+
sapiens_model.eval()
|
8 |
+
|
9 |
+
# Load MotionBERT model
|
10 |
+
motionbert_model = AutoModelForSequenceClassification.from_pretrained('/models/motionbert')
|
11 |
+
motionbert_tokenizer = AutoTokenizer.from_pretrained('/models/motionbert')
|
12 |
+
|
13 |
+
app = Flask(__name__)
|
14 |
+
|
15 |
+
@app.route('/pose_estimation', methods=['POST'])
|
16 |
+
def pose_estimation():
|
17 |
+
# Accept an image file as input for pose estimation
|
18 |
+
image = request.files['image'].read()
|
19 |
+
# Perform pose estimation
|
20 |
+
with torch.no_grad():
|
21 |
+
pose_result = sapiens_model(torch.tensor(image))
|
22 |
+
return jsonify({"pose_result": pose_result.tolist()})
|
23 |
+
|
24 |
+
@app.route('/sequence_analysis', methods=['POST'])
|
25 |
+
def sequence_analysis():
|
26 |
+
# Accept keypoint data as input for sequence analysis
|
27 |
+
keypoints = request.json['keypoints']
|
28 |
+
inputs = motionbert_tokenizer(keypoints, return_tensors="pt")
|
29 |
+
with torch.no_grad():
|
30 |
+
sequence_output = motionbert_model(**inputs)
|
31 |
+
return jsonify({"sequence_analysis": sequence_output.logits.tolist()})
|
32 |
+
|
33 |
+
if __name__ == '__main__':
|
34 |
+
app.run(host='0.0.0.0', port=7860)
|