app.py

from fastapi import FastAPI, UploadFile
from ultralytics import YOLOE
import io
from PIL import Image
import numpy as np
import os
from huggingface_hub import hf_hub_download
from ultralytics import YOLO
import requests
import supervision as sv

###

#pip install -q "git+https://github.com/THU-MIG/yoloe.git#subdirectory=third_party/CLIP"
#pip install -q "git+https://github.com/THU-MIG/yoloe.git#subdirectory=third_party/ml-mobileclip"
#pip install -q "git+https://github.com/THU-MIG/yoloe.git#subdirectory=third_party/lvis-api"
#pip install -q "git+https://github.com/THU-MIG/yoloe.git"

#wget -q https://docs-assets.developer.apple.com/ml-research/datasets/mobileclip/mobileclip_blt.pt


def init_model(model_id: str):
    is_pf=True
    # Create a models directory if it doesn't exist
    os.makedirs("models", exist_ok=True)
    filename = f"{model_id}-seg.pt" if not is_pf else f"{model_id}-seg-pf.pt"
    path = hf_hub_download(repo_id="jameslahm/yoloe", filename=filename)
    local_path = os.path.join("models", path)
    # Download and load model
    model = YOLOE(local_path)
    model.eval()
    return model

app = FastAPI()


@app.post("/predict")
async def predict(image_url: str,
                  texts: str = "hat",
                  model_id: str = "yoloe-11l",
                  conf: float = 0.25,
                  iou: float = 0.7
                  ):
    # Initialize model at startup
    model = init_model(model_id)
    # Set classes to filter
    class_list = [text.strip() for text in texts.split(',')]
    
    # Download and open image from URL
    response = requests.get(image_url)
    image = Image.open(io.BytesIO(response.content))
    
    # Get text embeddings and set classes properly
    text_embeddings = model.get_text_pe(class_list)
    model.set_classes(class_list, text_embeddings)
    
    # Run inference with the PIL Image
    results = model.predict(source=image, conf=conf, iou=iou)
    
    # Extract detection results
    result = results[0]
    # print(result)
    detections = []
    
    for box in result.boxes:
        detection = {
            "class": result.names[int(box.cls[0])],
            "confidence": float(box.conf[0]),
            "bbox": box.xyxy[0].tolist()  # Convert bbox tensor to list
        }
        detections.append(detection)
    print(detections)
    # detections = sv.Detections.from_ultralytics(results[0])

    # resolution_wh = image.size
    # thickness = sv.calculate_optimal_line_thickness(resolution_wh=resolution_wh)
    # text_scale = sv.calculate_optimal_text_scale(resolution_wh=resolution_wh)

    # labels = [
    #     f"{class_name} {confidence:.2f}"
    #     for class_name, confidence
    #     in zip(detections['class_name'], detections.confidence)
    # ]

    # annotated_image = image.copy()
    # annotated_image = sv.MaskAnnotator(color_lookup=sv.ColorLookup.INDEX, opacity=0.4).annotate(
    #     scene=annotated_image, detections=detections)
    # annotated_image = sv.BoxAnnotator(color_lookup=sv.ColorLookup.INDEX, thickness=thickness).annotate(
    #     scene=annotated_image, detections=detections)
    # annotated_image = sv.LabelAnnotator(color_lookup=sv.ColorLookup.INDEX, text_scale=text_scale, smart_position=True).annotate(
    #     scene=annotated_image, detections=detections, labels=labels)
    return {"detections": detections}

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)