# UVIS - Gradio App with Upload, URL & Video Support """ This script launches the UVIS (Unified Visual Intelligence System) as a Gradio Web App. Supports image, video, and URL-based media inputs for detection, segmentation, and depth estimation. Outputs include scene blueprint, structured JSON, and downloadable results. """ import time import logging import gradio as gr from PIL import Image import cv2 import timeout_decorator import spaces from registry import get_model from core.describe_scene import describe_scene from core.process import process_image from core.input_handler import resolve_input, validate_video, validate_image from utils.helpers import format_error, generate_session_id, toggle_visibility from huggingface_hub import hf_hub_download # Setup logging logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s') logger = logging.getLogger(__name__) # Model mappings DETECTION_MODEL_MAP = { "YOLOv5-Nano": "yolov5n-seg", "YOLOv5-Small": "yolov5s-seg", "YOLOv8-Small": "yolov8s", "YOLOv8-Large": "yolov8l", "RT-DETR": "rtdetr" # For future support } SEGMENTATION_MODEL_MAP = { "SegFormer-B0": "nvidia/segformer-b0-finetuned-ade-512-512", "SegFormer-B5": "nvidia/segformer-b5-finetuned-ade-512-512", "DeepLabV3-ResNet50": "deeplabv3_resnet50" } DEPTH_MODEL_MAP = { "MiDaS v21 Small 256": "midas_v21_small_256", "MiDaS v21 384": "midas_v21_384", "DPT Hybrid 384": "dpt_hybrid_384", "DPT Swin2 Large 384": "dpt_swin2_large_384", "DPT Beit Large 512": "dpt_beit_large_512" } # Resource Limits MAX_IMAGE_MB = 5 MAX_IMAGE_RES = (1920, 1080) MAX_VIDEO_MB = 50 MAX_VIDEO_DURATION = 30 # seconds @spaces.GPU def preload_models(): """ This function is needed to activate ZeroGPU. It must be decorated with @spaces.GPU. It can be used to warm up models or load them into memory. """ from registry import get_model print("Warming up models for ZeroGPU...") get_model("detection", "yolov5n-seg", device="cpu") get_model("segmentation", "deeplabv3_resnet50", device="cpu") get_model("depth", "midas_v21_small_256", device="cpu") # Main Handler def handle(mode, media_upload, url, run_det, det_model, det_confidence, run_seg, seg_model, run_depth, depth_model, blend): """ Master handler for resolving input and processing. Returns outputs for Gradio interface. """ session_id = generate_session_id() logger.info(f"Session ID: {session_id} | Handler activated with mode: {mode}") start_time = time.time() media = resolve_input(mode, media_upload, url) if not media: return None, format_error("No valid input provided. Please check your upload or URL."), None results = [] for single_media in media: if isinstance(single_media, str): # Video file valid, err = validate_video(single_media) if not valid: return None, format_error(err), None cap = cv2.VideoCapture(single_media) ret, frame = cap.read() cap.release() if not ret: return None, format_error("Failed to read video frame."), None single_media = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) if isinstance(single_media, Image.Image): valid, err = validate_image(single_media) if not valid: return None, format_error(err), None try: return process_image(single_media, run_det, det_model, det_confidence, run_seg, seg_model, run_depth, depth_model, blend) except timeout_decorator.timeout_decorator.TimeoutError: logger.error("Image processing timed out.") return None, format_error("Processing timed out. Try a smaller image or simpler model."), None logger.warning("Unsupported media type resolved.") log_runtime(start_time) return None, format_error("Invalid input. Please check your upload or URL."), None # Gradio Interface with gr.Blocks() as demo: gr.Markdown("## Unified Visual Intelligence System (UVIS)") with gr.Row(): with gr.Column(scale=2): # Input Mode Toggle mode = gr.Radio(["Upload", "URL"], value="Upload", label="Input Mode") # File upload: accepts multiple images or one video (user chooses wisely) media_upload = gr.File( label="Upload Images (1–5) or 1 Video", file_types=["image", ".mp4", ".mov", ".avi"], file_count="multiple" ) # URL input url = gr.Textbox(label="URL (Image/Video)", visible=False) # Toggle visibility def toggle_inputs(selected_mode): return [ gr.update(visible=(selected_mode == "Upload")), # media_upload gr.update(visible=(selected_mode == "URL")) # url ] mode.change(toggle_inputs, inputs=mode, outputs=[media_upload, url]) run_det = gr.Checkbox(label="Object Detection") run_seg = gr.Checkbox(label="Semantic Segmentation") run_depth = gr.Checkbox(label="Depth Estimation") det_model = gr.Dropdown(choices=list(DETECTION_MODEL_MAP), label="Detection Model", visible=False) seg_model = gr.Dropdown(choices=list(SEGMENTATION_MODEL_MAP), label="Segmentation Model", visible=False) depth_model = gr.Dropdown(choices=list(DEPTH_MODEL_MAP), label="Depth Model", visible=False) det_confidence = gr.Slider(0.1, 1.0, 0.5, label="Detection Confidence Threshold", visible=False) # Attach Visibility Logic run_det.change(toggle_visibility, inputs=[run_det], outputs=[det_model, det_confidence]) run_seg.change(toggle_visibility, inputs=[run_seg], outputs=[seg_model]) run_depth.change(toggle_visibility, inputs=[run_depth], outputs=[depth_model]) blend = gr.Slider(0.0, 1.0, 0.5, label="Overlay Blend") # Run Button run = gr.Button("Run Analysis") #Right panel with gr.Column(scale=1): # single_img_preview = gr.Image(label="Preview (Image)", visible=False) # gallery_preview = gr.Gallery(label="Preview (Gallery)", columns=3, height="auto", visible=False) # video_preview = gr.Video(label="Preview (Video)", visible=False) img_out = gr.Image(label="Scene Blueprint") json_out = gr.JSON(label="Scene JSON") zip_out = gr.File(label="Download Results") # # Output Tabs # with gr.Tab("Scene JSON"): # json_out = gr.JSON() # with gr.Tab("Scene Blueprint"): # img_out = gr.Image() # with gr.Tab("Download"): # zip_out = gr.File() # Button Click Event run.click( handle, inputs=[mode, media_upload, url, run_det, det_model, det_confidence, run_seg, seg_model, run_depth, depth_model, blend], outputs=[img_out, json_out, zip_out] ) # Footer Section gr.Markdown("---") gr.Markdown( """
Built by Durga Deepak Valluri
GitHub | Website | LinkedIn
""", ) # Launch the Gradio App demo.launch()