Spaces:
Sleeping
Sleeping
File size: 6,479 Bytes
f137487 867f506 f137487 867f506 f137487 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 |
from fastapi import FastAPI, UploadFile, File
import cv2
import torch
import pandas as pd
from PIL import Image
from transformers import AutoImageProcessor, AutoModelForImageClassification
from tqdm import tqdm
import json
import shutil
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import HTMLResponse
app = FastAPI()
# Add CORS middleware to allow requests from localhost:8080 (or any origin you specify)
app.add_middleware(
CORSMiddleware,
# allow_origins=["http://localhost:8080"], # Replace with the URL of your Vue.js app
allow_origins=["http://localhost:8080"], # Replace with the URL of your Vue.js app
allow_credentials=True,
allow_methods=["*"], # Allows all HTTP methods (GET, POST, etc.)
allow_headers=["*"], # Allows all headers (such as Content-Type, Authorization, etc.)
)
# Charger le processor et le modèle fine-tuné depuis le chemin local
local_model_path = r'.\vit-finetuned-ucf101'
processor = AutoImageProcessor.from_pretrained("google/vit-base-patch16-224")
model = AutoModelForImageClassification.from_pretrained(local_model_path)
# model = AutoModelForImageClassification.from_pretrained("2nzi/vit-finetuned-ucf101")
model.eval()
# Fonction pour classifier une image
def classifier_image(image):
image_pil = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
inputs = processor(images=image_pil, return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
logits = outputs.logits
predicted_class_idx = logits.argmax(-1).item()
predicted_class = model.config.id2label[predicted_class_idx]
return predicted_class
# Fonction pour traiter la vidéo et identifier les séquences de "Surfing"
def identifier_sequences_surfing(video_path, intervalle=0.5):
cap = cv2.VideoCapture(video_path)
frame_rate = cap.get(cv2.CAP_PROP_FPS)
total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
frame_interval = int(frame_rate * intervalle)
resultats = []
sequences_surfing = []
frame_index = 0
in_surf_sequence = False
start_timestamp = None
with tqdm(total=total_frames, desc="Traitement des frames de la vidéo", unit="frame") as pbar:
success, frame = cap.read()
while success:
if frame_index % frame_interval == 0:
timestamp = round(frame_index / frame_rate, 2) # Maintain precision to the centisecond level
classe = classifier_image(frame)
resultats.append({"Timestamp": timestamp, "Classe": classe})
if classe == "Surfing" and not in_surf_sequence:
in_surf_sequence = True
start_timestamp = timestamp
elif classe != "Surfing" and in_surf_sequence:
# Vérifier l'image suivante pour confirmer si c'était une erreur ponctuelle
success_next, frame_next = cap.read()
next_timestamp = round((frame_index + frame_interval) / frame_rate, 2)
classe_next = None
if success_next:
classe_next = classifier_image(frame_next)
resultats.append({"Timestamp": next_timestamp, "Classe": classe_next})
# Si l'image suivante est "Surfing", on ignore l'erreur ponctuelle
if classe_next == "Surfing":
success = success_next
frame = frame_next
frame_index += frame_interval
pbar.update(frame_interval)
continue
else:
# Sinon, terminer la séquence "Surfing"
in_surf_sequence = False
end_timestamp = timestamp
sequences_surfing.append((start_timestamp, end_timestamp))
success, frame = cap.read()
frame_index += 1
pbar.update(1)
# Si on est toujours dans une séquence "Surfing" à la fin de la vidéo
if in_surf_sequence:
sequences_surfing.append((start_timestamp, round(frame_index / frame_rate, 2)))
cap.release()
dataframe_sequences = pd.DataFrame(sequences_surfing, columns=["Début", "Fin"])
return dataframe_sequences
# Fonction pour convertir les séquences en format JSON
def convertir_sequences_en_json(dataframe):
events = []
blocks = []
for idx, row in dataframe.iterrows():
block = {
"id": f"Surfing{idx + 1}",
"start": round(row["Début"], 2),
"end": round(row["Fin"], 2)
}
blocks.append(block)
event = {
"event": "Surfing",
"blocks": blocks
}
events.append(event)
return events
@app.post("/analyze_video/")
async def analyze_video(file: UploadFile = File(...)):
with open("uploaded_video.mp4", "wb") as buffer:
shutil.copyfileobj(file.file, buffer)
dataframe_sequences = identifier_sequences_surfing("uploaded_video.mp4", intervalle=1)
json_result = convertir_sequences_en_json(dataframe_sequences)
return json_result
@app.get("/", response_class=HTMLResponse)
async def index():
return (
"""
<html>
<body>
<h1>Hello world!</h1>
<p>This `/` is the most simple and default endpoint.</p>
<p>If you want to learn more, check out the documentation of the API at
<a href='/docs'>/docs</a> or
<a href='https://2nzi-video-sequence-labeling.hf.space/docs' target='_blank'>external docs</a>.
</p>
</body>
</html>
"""
)
# Lancer l'application avec uvicorn (command line)
# uvicorn main:app --reload
# http://localhost:8000/docs#/
# (.venv) PS C:\Users\antoi\Documents\Work_Learn\Labeling-Deploy\FastAPI> uvicorn main:app --host 0.0.0.0 --port 8000 --workers 1 |