app.py

# STEP 1: Install dependencies
# Note: Use requirements.txt when deploying
import torch
from transformers import AutoImageProcessor, AutoModelForObjectDetection
from PIL import Image, ImageDraw, ImageFont
import gradio as gr

# STEP 2: Load YOLOS model & processor
model_name = "hustvl/yolos-base"
processor = AutoImageProcessor.from_pretrained(model_name)
model = AutoModelForObjectDetection.from_pretrained(model_name)
model.eval()

if torch.cuda.is_available():
    model.to(torch.float16).to("cuda")

# STEP 3: Detection function with object name return
def detect_yolos(image, threshold=0.5):
    image = image.convert("RGB")
    inputs = processor(images=image, return_tensors="pt").to(model.device)
    with torch.no_grad():
        outputs = model(**inputs)

    target_sizes = torch.tensor([image.size[::-1]], device=model.device)
    results = processor.post_process_object_detection(outputs, threshold=threshold, target_sizes=target_sizes)[0]

    draw = ImageDraw.Draw(image)
    font = ImageFont.load_default()
    detected_labels = []

    for score, label_idx, box in zip(results["scores"], results["labels"], results["boxes"]):
        label = model.config.id2label[label_idx.item()]
        detected_labels.append(label)
        box = [round(i, 2) for i in box.tolist()]
        draw.rectangle(box, outline="green", width=2)
        draw.text((box[0], box[1] - 10), f"{label}: {score:.2f}", fill="green", font=font)

    label_summary = ", ".join(set(detected_labels)) if detected_labels else "No objects detected."
    return image, label_summary

# STEP 4: Gradio UI
demo = gr.Interface(
    fn=detect_yolos,
    inputs=[
        gr.Image(type="pil", label="Upload Image"),
        gr.Slider(0, 1, value=0.5, label="Confidence Threshold")
    ],
    outputs=[
        gr.Image(type="pil", label="Image with Detections"),
        gr.Textbox(label="Detected Object Names")
    ],
    title="📦 YOLOS Object Detection + Label List",
    description="Detects objects using YOLOS and lists all object names in a textbox."
)

demo.launch()