app.py

import os
import cv2
import tqdm
import uuid
import logging

import torch
import trackers
import numpy as np
import gradio as gr
import imageio.v3 as iio
import supervision as sv

from pathlib import Path
from typing import List, Optional, Tuple

from PIL import Image
from pipeline import build_pipeline
from utils import cfg, load_config, load_onnx_model


# Configuration constants
DETECTORS = {
    "yolo8n-640": 'downloads/yolo8n-640.onnx',
    "yolo8n-416": 'downloads/yolo8n-416.onnx',
}
DEFAULT_DETECTOR = list(DETECTORS.keys())[0]
DEFAULT_CONFIDENCE_THRESHOLD = 0.6


# Image
IMAGE_EXAMPLES = [
    {"path": "./examples/images/forest.jpg", "label": "Local Image"},
    {"path": "./examples/images/river.jpg", "label": "Local Image"},
    {"path": "./examples/images/ocean.jpg", "label": "Local Image"},
]

# Video
MAX_NUM_FRAMES = 250
ALLOWED_VIDEO_EXTENSIONS = {".mp4", ".avi", ".mov"}
VIDEO_OUTPUT_DIR = Path("static/videos")
VIDEO_OUTPUT_DIR.mkdir(parents=True, exist_ok=True)

class TrackingAlgorithm:
    BYTETRACK = "ByteTrack (2021)"
    DEEPSORT = "DeepSORT (2017)"
    SORT = "SORT (2016)"

TRACKERS = [None, TrackingAlgorithm.BYTETRACK, TrackingAlgorithm.DEEPSORT, TrackingAlgorithm.SORT]
VIDEO_EXAMPLES = [
    {"path": "./examples/videos/sea.mp4", "label": "Local Video", "tracker": TrackingAlgorithm.BYTETRACK, "classes": "Person, Boat"},
    {"path": "./examples/videos/forest.mp4", "label": "Local Video", "tracker": TrackingAlgorithm.BYTETRACK, "classes": "LightVehicle, Person, Boat"},
]


# Create a color palette for visualization
# These hex color codes define different colors for tracking different objects
color = sv.ColorPalette.from_hex([
    "#ffff00", "#ff9b00", "#ff8080", "#ff66b2", "#ff66ff", "#b266ff",
    "#9999ff", "#3399ff", "#66ffff", "#33ff99", "#66ff66", "#99ff00"
])


logging.basicConfig(
    level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s"
)
logger = logging.getLogger(__name__)


def get_pipeline(config: dict, onnx_path: str):
    pipeline = build_pipeline(config)
    load_onnx_model(pipeline.detector, onnx_path) 
    return pipeline


def detect_objects(
    config: dict,
    onnx_path: str,
    images: List[np.ndarray] | np.ndarray, 
    confidence_threshold: float = DEFAULT_CONFIDENCE_THRESHOLD,
    target_size: Optional[Tuple[int, int]] = None,
    classes: Optional[List[str]] = None,
):
    config.defrost()
    config.detector.thresholds.confidence = float(confidence_threshold)
    config.freeze()
    pipeline = get_pipeline(config, onnx_path)
    id2label = pipeline.detector.get_category_mapping()
    label2id = {v: k for k, v in pipeline.detector.get_category_mapping().items()}
    if classes is not None:
        wrong_classes = [cls for cls in classes if cls not in label2id]
        if wrong_classes:
            gr.Warning(f"Classes not found in model config: {wrong_classes}")
        keep_ids = [label2id[cls] for cls in classes if cls in label2id]
    else:
        keep_ids = None

    if isinstance(images, np.ndarray) and images.ndim == 4:
        images = [x for x in images]  # split video array into list of images

    results = []
    for img in tqdm.tqdm(images, desc="Processing frames"):
        output_ = pipeline(img)
        output_reshaped = {
        "scores": torch.from_numpy(output_.confidence) if isinstance(output_.confidence, np.ndarray) else output_.confidence,
        "labels": torch.from_numpy(output_.class_id) if isinstance(output_.class_id, np.ndarray) else output_.class_id,
        "boxes": torch.from_numpy(output_.xyxy) if isinstance(output_.xyxy, np.ndarray) else output_.xyxy,
        }
        results.append(output_reshaped)
        if target_size:
            # Resize boxes to target size
            scale_x = target_size[0] / img.shape[1]
            scale_y = target_size[1] / img.shape[0]
            output_reshaped["boxes"][:, [0, 2]] *= scale_x
            output_reshaped["boxes"][:, [1, 3]] *= scale_y
        
        

    # # move results to cpu
    for i, result in enumerate(results):
        results[i] = {k: v for k, v in result.items()}
        if keep_ids is not None:
            keep = torch.isin(results[i]["labels"], torch.tensor(keep_ids))
            results[i] = {k: v[keep] for k, v in results[i].items()}
    
    # return results, model.config.id2label
    return results, pipeline.detector.get_category_mapping()


def process_image(
    model: str = DEFAULT_DETECTOR,
    image: Optional[Image.Image] = None,
    confidence_threshold: float = DEFAULT_CONFIDENCE_THRESHOLD,
):
    
    load_config(cfg, f'configs/{model}.yaml')
    results, id2label = detect_objects(
        config=cfg.pipeline,
        onnx_path=DETECTORS[model],
        images=[np.array(image)],
        confidence_threshold=confidence_threshold,
    )
    result = results[0] # first image in batch (we have batch size 1)

    annotations = []
    for label, score, box in zip(result["labels"], result["scores"], result["boxes"]):
        text_label = id2label[label.item()]
        formatted_label = f"{text_label} ({score:.2f})"
        x_min, y_min, x_max, y_max = box.cpu().numpy().round().astype(int)
        x_min = max(0, x_min)
        y_min = max(0, y_min)
        x_max = min(image.width - 1, x_max)
        y_max = min(image.height - 1, y_max)
        annotations.append(((x_min, y_min, x_max, y_max), formatted_label))

    return (image, annotations)


def get_target_size(image_height, image_width, max_size: int):
    if image_height < max_size and image_width < max_size:
        new_height, new_width = image_height, image_width 
    elif image_height > image_width:
        new_height = max_size
        new_width = int(image_width * max_size / image_height)
    else:
        new_width = max_size
        new_height = int(image_height * max_size / image_width)
    
    # make even (for video codec compatibility)
    new_height = new_height // 2 * 2
    new_width = new_width // 2 * 2

    return new_width, new_height


def read_video_k_frames(video_path: str, k: int, read_every_i_frame: int = 1):
    cap = cv2.VideoCapture(video_path)
    frames = []
    i = 0
    progress_bar = tqdm.tqdm(total=k, desc="Reading frames")
    while cap.isOpened() and len(frames) < k:
        ret, frame = cap.read()
        if not ret:
            break
        if i % read_every_i_frame == 0:
            frames.append(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
            progress_bar.update(1)
        i += 1
    cap.release()
    progress_bar.close()
    return frames


def get_tracker(tracker: str, fps: float):
    if tracker == TrackingAlgorithm.SORT:
        return trackers.SORTTracker(frame_rate=fps)
    elif tracker == TrackingAlgorithm.DEEPSORT:
        feature_extractor = trackers.DeepSORTFeatureExtractor.from_timm("mobilenetv4_conv_small.e1200_r224_in1k", device="cpu")
        return trackers.DeepSORTTracker(feature_extractor, frame_rate=fps)
    elif tracker == TrackingAlgorithm.BYTETRACK:  
        return sv.ByteTrack(frame_rate=int(fps))
    else:
        raise ValueError(f"Invalid tracker: {tracker}")


def update_tracker(tracker, detections, frame):
    tracker_name = tracker.__class__.__name__
    if tracker_name == "SORTTracker":
        return tracker.update(detections)
    elif tracker_name == "DeepSORTTracker":
        return tracker.update(detections, frame)
    elif tracker_name == "ByteTrack":
        return tracker.update_with_detections(detections)
    else:
        raise ValueError(f"Invalid tracker: {tracker}")


def process_video(
    video_path: str,
    checkpoint: str,
    tracker_algorithm: Optional[str] = None,
    classes: str = "all",
    confidence_threshold: float = DEFAULT_CONFIDENCE_THRESHOLD,
    progress: gr.Progress = gr.Progress(track_tqdm=True),
) -> str:
    
    if not video_path or not os.path.isfile(video_path):
        raise ValueError(f"Invalid video path: {video_path}")

    ext = os.path.splitext(video_path)[1].lower()
    if ext not in ALLOWED_VIDEO_EXTENSIONS:
        raise ValueError(f"Unsupported video format: {ext}, supported formats: {ALLOWED_VIDEO_EXTENSIONS}")

    video_info = sv.VideoInfo.from_video_path(video_path)
    read_each_i_frame = max(1, video_info.fps // 25)
    target_fps = video_info.fps / read_each_i_frame
    target_width, target_height = get_target_size(video_info.height, video_info.width, 1080)

    n_frames_to_read = min(MAX_NUM_FRAMES, video_info.total_frames // read_each_i_frame)
    frames = read_video_k_frames(video_path, n_frames_to_read, read_each_i_frame)
    frames = [cv2.resize(frame, (target_width, target_height), interpolation=cv2.INTER_CUBIC) for frame in frames]

    # Set the color lookup mode to assign colors by track ID
    # This mean objects with the same track ID will be annotated by the same color
    color_lookup = sv.ColorLookup.TRACK if tracker_algorithm else sv.ColorLookup.CLASS

    box_annotator = sv.BoxAnnotator(color, color_lookup=color_lookup, thickness=1)
    label_annotator = sv.LabelAnnotator(color, color_lookup=color_lookup, text_scale=0.5)

    # preprocess classes
    if classes != "all":
        classes_list = [cls.strip() for cls in classes.split(",")]
    else:
        classes_list = None

    load_config(cfg, f'configs/{checkpoint}.yaml')
    results, id2label = detect_objects(
        config=cfg.pipeline,
        onnx_path=DETECTORS[checkpoint],
        images=np.array(frames),
        confidence_threshold=confidence_threshold,
        target_size=(target_height, target_width),
        classes=classes_list,
    )


    annotated_frames = []

    # detections
    if tracker_algorithm:
        tracker = get_tracker(tracker_algorithm, target_fps)
        for frame, result in progress.tqdm(zip(frames, results), desc="Tracking objects", total=len(frames)):
            detections = sv.Detections.from_transformers(result, id2label=id2label)
            detections = detections.with_nms(threshold=0.95, class_agnostic=True)
            detections = update_tracker(tracker, detections, frame)
            labels = [f"#{tracker_id} {id2label[class_id]}" for class_id, tracker_id in zip(detections.class_id, detections.tracker_id)]
            annotated_frame = box_annotator.annotate(scene=frame, detections=detections)
            annotated_frame = label_annotator.annotate(scene=annotated_frame, detections=detections, labels=labels)
            annotated_frames.append(annotated_frame)
        
    else:
        for frame, result in tqdm.tqdm(zip(frames, results), desc="Annotating frames", total=len(frames)):
            detections = sv.Detections.from_transformers(result, id2label=id2label)
            detections = detections.with_nms(threshold=0.95, class_agnostic=True)
            annotated_frame = box_annotator.annotate(scene=frame, detections=detections)
            annotated_frame = label_annotator.annotate(scene=annotated_frame, detections=detections)
            annotated_frames.append(annotated_frame)

    output_filename = os.path.join(VIDEO_OUTPUT_DIR, f"output_{uuid.uuid4()}.mp4")
    iio.imwrite(output_filename, annotated_frames, fps=target_fps, codec="h264")
    return output_filename



def create_image_inputs() -> List[gr.components.Component]:
    return [
        gr.Image(
            label="Upload Image",
            type="pil",
            sources=["upload", "webcam"],
            interactive=True,
            elem_classes="input-component",
        ),
        gr.Dropdown(
            choices=list(DETECTORS.keys()),
            label="Select Model Checkpoint",
            value=DEFAULT_DETECTOR,
            elem_classes="input-component",
        ),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=DEFAULT_CONFIDENCE_THRESHOLD,
            step=0.1,
            label="Confidence Threshold",
            elem_classes="input-component",
        ),
    ]


def create_video_inputs() -> List[gr.components.Component]:
    return [
        gr.Video(
            label="Upload Video",
            sources=["upload"],
            interactive=True,
            format="mp4",  # Ensure MP4 format
            elem_classes="input-component",
        ),
        gr.Dropdown(
            choices=list(DETECTORS.keys()),
            label="Select Model Checkpoint",
            value=DEFAULT_DETECTOR,
            elem_classes="input-component",
        ),
        gr.Dropdown(
            choices=TRACKERS,
            label="Select Tracker (Optional)",
            value=None,
            elem_classes="input-component",
        ),
        gr.TextArea(
            label="Specify Class Names to Detect (comma separated)",
            value="all",
            lines=1,
            elem_classes="input-component",
        ),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=DEFAULT_CONFIDENCE_THRESHOLD,
            step=0.1,
            label="Confidence Threshold",
            elem_classes="input-component",
        ),
    ]


def create_button_row() -> List[gr.Button]:
    return [
        gr.Button(
            f"Detect Objects", variant="primary", elem_classes="action-button"
        ),
        gr.Button(f"Clear", variant="secondary", elem_classes="action-button"),
    ]


# Gradio interface
with gr.Blocks(theme=gr.themes.Ocean()) as demo:
    gr.Markdown(
        """
        # Pipeline for Aerial Search and Rescue Demo
        Experience state-of-the-art object detection with Open Source [WALDO30](https://huggingface.co/StephanST/WALDO30) models.
         - **Image** and **Video** modes are supported.
         - Select a model and adjust the confidence threshold to see detections!
         - On video mode, you can enable tracking powered by [Supervision](https://github.com/roboflow/supervision) and [Trackers](https://github.com/roboflow/trackers) from Roboflow.

        For more details and source code, visit the [PiSAR](https://github.com/eadali/PiSAR).
        """,
        elem_classes="header-text",
    )

    with gr.Tabs():
        with gr.Tab("Image"):
            with gr.Row():
                with gr.Column(scale=1, min_width=300):
                    with gr.Group():
                        (
                            image_input,
                            image_model_checkpoint,
                            image_confidence_threshold,
                        ) = create_image_inputs()
                        image_detect_button, image_clear_button = create_button_row()
                with gr.Column(scale=2):
                    image_output = gr.AnnotatedImage(
                        label="Detection Results",
                        show_label=True,
                        color_map=None,
                        elem_classes="output-component",
                    )
            gr.Examples(
                examples=[
                    [
                        DEFAULT_DETECTOR,
                        example["path"],
                        DEFAULT_CONFIDENCE_THRESHOLD,
                    ]
                    for example in IMAGE_EXAMPLES
                ],
                inputs=[
                    image_model_checkpoint,
                    image_input,
                    image_confidence_threshold,
                ],
                outputs=[image_output],
                fn=process_image,
                label="Select an image example to populate inputs",
                cache_examples=True,
                cache_mode="lazy",
            )

        with gr.Tab("Video"):
            gr.Markdown(
                f"The input video will be processed in ~25 FPS (up to {MAX_NUM_FRAMES} frames in result)."
            )
            with gr.Row():
                with gr.Column(scale=1, min_width=300):
                    with gr.Group():
                        video_input, video_checkpoint, video_tracker, video_classes, video_confidence_threshold = create_video_inputs()
                        video_detect_button, video_clear_button = create_button_row()
                with gr.Column(scale=2):
                    video_output = gr.Video(
                        label="Detection Results",
                        format="mp4",  # Explicit MP4 format
                        elem_classes="output-component",
                    )

            gr.Examples(
                examples=[
                    [example["path"], DEFAULT_DETECTOR, example["tracker"], example["classes"], DEFAULT_CONFIDENCE_THRESHOLD]
                    for example in VIDEO_EXAMPLES
                ],
                inputs=[video_input, video_checkpoint, video_tracker, video_classes, video_confidence_threshold],
                outputs=[video_output],
                fn=process_video,
                cache_examples=False,
                label="Select a video example to populate inputs",
            )

    # Image clear button
    image_clear_button.click(
        fn=lambda: (
            None,
            DEFAULT_DETECTOR,
            DEFAULT_CONFIDENCE_THRESHOLD,
            None,
        ),
        outputs=[
            image_input,
            image_model_checkpoint,
            image_confidence_threshold,
            image_output,
        ],
    )

    # Video clear button
    video_clear_button.click(
        fn=lambda: (
            None,
            DEFAULT_DETECTOR,
            None,
            "all",
            DEFAULT_CONFIDENCE_THRESHOLD,
            None,
        ),
        outputs=[
            video_input,
            video_checkpoint,
            video_tracker,
            video_classes,
            video_confidence_threshold,
            video_output,
        ],
    )

    # Image detect button
    image_detect_button.click(
        fn=process_image,
        inputs=[
            image_model_checkpoint,
            image_input,
            image_confidence_threshold,
        ],
        outputs=[image_output],
    )

    # Video detect button
    video_detect_button.click(
        fn=process_video,
        inputs=[video_input, video_checkpoint, video_tracker, video_classes, video_confidence_threshold],
        outputs=[video_output],
    )

if __name__ == "__main__":
    demo.queue(max_size=20).launch()