app.py

#!/usr/bin/env python3
"""
CropScan - Plant Disease Detection
Hugging Face Space Demo
"""

import gradio as gr
import numpy as np
from PIL import Image
from pathlib import Path
import torch

# Model paths
MODEL_DIR = Path("models")
RFDETR_CHECKPOINT = MODEL_DIR / "rfdetr" / "checkpoint_best_total.pth"
SAM2_CHECKPOINT = MODEL_DIR / "sam2" / "sam2.1_hiera_small.pt"

# Lazy loaded components
segmenter = None
leaf_segmenter = None

# Clean CSS - green theme, no effects
CUSTOM_CSS = """
.gradio-container {
    background: #0d1117 !important;
    font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
}

.main-header {
    text-align: center;
    padding: 1.5rem 0;
    border-bottom: 1px solid #21262d;
    margin-bottom: 1.5rem;
}

.main-header h1 {
    font-size: 2rem;
    font-weight: 600;
    color: #3fb950;
    margin: 0;
}

.main-header p {
    color: #8b949e;
    margin-top: 0.25rem;
    font-size: 0.9rem;
}

.section-label {
    color: #c9d1d9;
    font-size: 0.875rem;
    font-weight: 500;
    margin-bottom: 0.75rem;
}

.result-box {
    background: #161b22;
    border: 1px solid #30363d;
    border-radius: 8px;
    padding: 1rem;
    color: #c9d1d9;
}

.healthy-status {
    background: #0d1117;
    border: 1px solid #238636;
    border-radius: 6px;
    padding: 1rem;
    margin-bottom: 1rem;
}

.healthy-status h4 {
    color: #3fb950;
    margin: 0 0 0.25rem 0;
    font-size: 1rem;
}

.warning-status {
    background: #0d1117;
    border: 1px solid #9e6a03;
    border-radius: 6px;
    padding: 1rem;
    margin-bottom: 1rem;
}

.warning-status h4 {
    color: #d29922;
    margin: 0 0 0.25rem 0;
    font-size: 1rem;
}

.danger-status {
    background: #0d1117;
    border: 1px solid #da3633;
    border-radius: 6px;
    padding: 1rem;
    margin-bottom: 1rem;
}

.danger-status h4 {
    color: #f85149;
    margin: 0 0 0.25rem 0;
    font-size: 1rem;
}

button.primary {
    background: #238636 !important;
    border: none !important;
}

button.primary:hover {
    background: #2ea043 !important;
}

input[type="range"] {
    accent-color: #238636 !important;
}

input[type="checkbox"] {
    accent-color: #238636 !important;
}

.footer-text {
    text-align: center;
    color: #484f58;
    font-size: 0.8rem;
    padding: 1rem;
    border-top: 1px solid #21262d;
    margin-top: 1rem;
}

.examples-title {
    color: #3fb950 !important;
    font-size: 1.1rem !important;
    font-weight: 600 !important;
    margin-bottom: 1rem !important;
}

.gallery img {
    height: 120px !important;
    width: auto !important;
    object-fit: cover !important;
}

footer {
    display: none !important;
}
"""


def load_models():
    """Load models on first use."""
    global segmenter, leaf_segmenter

    if segmenter is not None:
        return

    print("Loading models...")

    from src.sam3_segmentation import RFDETRSegmenter
    from src.leaf_segmenter import SAM2LeafSegmenter

    segmenter = RFDETRSegmenter(
        checkpoint_path=str(RFDETR_CHECKPOINT),
        model_size="medium"
    )

    leaf_segmenter = SAM2LeafSegmenter(
        checkpoint_path=str(SAM2_CHECKPOINT)
    )

    print("Models loaded!")


def get_care_recommendations(num_detections: int, affected_percent: float) -> str:
    """Generate care recommendations based on detection results."""

    if num_detections == 0:
        return """<div class="healthy-status">
<h4>Healthy</h4>
<p style="color: #8b949e; margin: 0;">No disease symptoms detected.</p>
</div>

**Care tips:**
- Continue regular watering
- Ensure adequate sunlight
- Monitor for changes
"""

    if affected_percent < 10:
        severity = "Low"
        status_class = "warning-status"
    elif affected_percent < 30:
        severity = "Moderate"
        status_class = "warning-status"
    else:
        severity = "High"
        status_class = "danger-status"

    return f"""<div class="{status_class}">
<h4>Disease Detected - {severity}</h4>
<p style="color: #8b949e; margin: 0;">{affected_percent:.1f}% affected | {num_detections} region(s)</p>
</div>

**Recommended actions:**

1. **Isolate** - Separate from healthy plants

2. **Remove affected leaves** - Prune with sterilized tools

3. **Treatment**
   - Copper-based fungicide
   - Neem oil spray
   - Improve air circulation

4. **Monitor** - Check daily for 1-2 weeks
"""


def detect_disease(
    image: np.ndarray,
    use_leaf_segmentation: bool = True,
    confidence_threshold: float = 0.3
) -> tuple:
    """Detect plant diseases in an image."""

    if image is None:
        return None, "Upload an image to start."

    load_models()

    pil_image = Image.fromarray(image)
    original_image = image.copy()

    segmented_image = None
    leaf_mask = None
    if use_leaf_segmentation:
        segmented_pil, leaf_mask = leaf_segmenter.auto_segment_leaf(
            pil_image, return_mask=True
        )
        segmented_image = np.array(segmented_pil)
        detection_input = segmented_pil
    else:
        detection_input = pil_image

    prompts = ["diseased plant tissue", "leaf spot", "disease symptom"]
    seg_result = segmenter.segment_with_concepts(
        detection_input,
        prompts,
        confidence_threshold=confidence_threshold
    )

    num_detections = len(seg_result.boxes)

    if num_detections > 0:
        refined_masks = leaf_segmenter.refine_boxes_to_masks(
            detection_input,
            seg_result.boxes
        )
    else:
        refined_masks = np.zeros((0, image.shape[0], image.shape[1]), dtype=bool)

    from src.visualization import create_mask_overlay

    if use_leaf_segmentation and segmented_image is not None:
        base_image = segmented_image
    else:
        base_image = original_image

    if num_detections > 0:
        annotated = create_mask_overlay(base_image, refined_masks, alpha=0.5)
    else:
        annotated = base_image

    affected_percent = 0
    if num_detections > 0:
        total_mask = np.zeros(refined_masks[0].shape, dtype=bool)
        for mask in refined_masks:
            total_mask |= mask

        if use_leaf_segmentation and leaf_mask is not None:
            affected_percent = (total_mask & leaf_mask).sum() / max(leaf_mask.sum(), 1) * 100
        else:
            affected_percent = total_mask.sum() / (image.shape[0] * image.shape[1]) * 100

    recommendations = get_care_recommendations(num_detections, affected_percent)

    return annotated, recommendations


def create_demo():
    """Create Gradio interface."""

    with gr.Blocks(
        title="CropScan",
        css=CUSTOM_CSS,
        theme=gr.themes.Base(
            primary_hue="green",
            neutral_hue="gray",
        ).set(
            body_background_fill="#0d1117",
            block_background_fill="#161b22",
            block_border_width="1px",
            block_border_color="#30363d",
            input_background_fill="#0d1117",
            button_primary_background_fill="#238636",
        )
    ) as demo:

        gr.HTML("""
        <div class="main-header">
            <h1>CropScan</h1>
            <p>Plant disease detection</p>
        </div>
        """)

        with gr.Row():
            with gr.Column():
                input_image = gr.Image(
                    label="Upload",
                    type="numpy",
                    height=350,
                    sources=["upload", "webcam"]
                )

                confidence_slider = gr.Slider(
                    minimum=0.1,
                    maximum=0.9,
                    value=0.3,
                    step=0.05,
                    label="Sensitivity"
                )

                leaf_seg_checkbox = gr.Checkbox(
                    value=False,
                    label="SAM2 precision mode"
                )

                detect_btn = gr.Button("Scan", variant="primary", size="lg")

            with gr.Column():
                output_image = gr.Image(
                    label="Result",
                    type="numpy",
                    height=350
                )

                detection_info = gr.Markdown()

                gr.HTML('<p class="examples-title">Examples - click to load</p>')

                example_images = [
                    ["img1.jpg"],
                    ["img2.jpg"],
                    ["img3.jpg"],
                    ["img4.jpg"],
                    ["img5.jpg"],
                ]

                gr.Examples(
                    examples=example_images,
                    inputs=[input_image],
                    examples_per_page=5,
                )

        detect_btn.click(
            fn=detect_disease,
            inputs=[input_image, leaf_seg_checkbox, confidence_slider],
            outputs=[output_image, detection_info]
        )

        gr.HTML('<div class="footer-text">RF-DETR + SAM2</div>')

    return demo


if __name__ == "__main__":
    demo = create_demo()
    demo.launch()