analysis/failure_case_analysis/ov_lvis/visualize_comparison.py

#!/usr/bin/env python3
"""
OV-LVIS Failure Case Analysis - Visualization (Rare Categories)

生成 rare 类别的可视化结果（DeCLIP vs CLIPSelf）：
- 直接使用分析结果中记录的bbox
- 清晰显示分类结果（正确/错误/未定位）

Usage:
    python visualize_comparison.py \
        --cases analysis_output/comparison_vitb16/top_advantage_cases.json \
        --ann-file /path/to/lvis_v1_val.json \
        --img-dir /path/to/lvis_v1 \
        --output analysis_output/visualizations/advantage \
        --case-type advantage
"""

import argparse
import json
from collections import defaultdict
from pathlib import Path
from typing import List, Optional

import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.patches as patches
import numpy as np
from PIL import Image


COLORS = {
    'correct': '#27AE60',
    'incorrect': '#E74C3C',
    'missed': '#95A5A6',
    'background': '#FFFFFF',
}

MODEL_NAMES = {
    'declip': 'DeCLIP (Ours)',
    'clipself': 'CLIPSelf',
}

MODEL_ORDER = ['declip', 'clipself']

plt.rcParams.update({
    'font.family': 'sans-serif',
    'font.sans-serif': ['Arial', 'DejaVu Sans', 'Helvetica', 'Liberation Sans'],
    'font.size': 10,
    'axes.titlesize': 12,
    'axes.labelsize': 10,
    'xtick.labelsize': 9,
    'ytick.labelsize': 9,
    'legend.fontsize': 9,
    'figure.titlesize': 14,
    'axes.linewidth': 0.8,
    'axes.edgecolor': '#CCCCCC',
})


def parse_args():
    parser = argparse.ArgumentParser(description='OV-LVIS visualization for rare analysis')
    parser.add_argument('--cases', required=True, help='Top cases JSON file (from compare_models_rare.py)')
    parser.add_argument('--ann-file', required=True, help='LVIS annotation file')
    parser.add_argument('--img-dir', required=True, help='Image directory (lvis_v1 root)')
    parser.add_argument('--output', required=True, help='Output directory')
    parser.add_argument('--dpi', type=int, default=150, help='Output DPI')
    parser.add_argument('--case-type', choices=['advantage', 'unsolved'], default='advantage',
                        help='Type of cases to visualize')
    return parser.parse_args()


def load_image_map(ann_file: str) -> dict:
    with open(ann_file, 'r') as f:
        data = json.load(f)
    return {img['id']: img['file_name'] for img in data['images']}


def draw_bbox(
    ax: plt.Axes,
    bbox: List[float],
    label: Optional[str],
    color: str,
    score: Optional[float] = None,
    linewidth: float = 2.5,
    alpha: float = 1.0,
    zorder: int = 2
):
    x, y, w, h = bbox
    rect = patches.Rectangle(
        (x, y), w, h,
        linewidth=linewidth,
        edgecolor=color,
        facecolor='none',
        alpha=alpha,
        zorder=zorder
    )
    ax.add_patch(rect)

    if not label:
        return

    if score is not None:
        text = f"{label} {score:.2f}"
    else:
        text = label

    fontsize = 8
    text_y = y - 3 if y > 20 else y + h + 12

    ax.text(
        x, text_y,
        text,
        fontsize=fontsize,
        fontweight='bold',
        color='white',
        bbox=dict(
            facecolor=color,
            edgecolor='none',
            alpha=0.85,
            pad=1.5,
            boxstyle='round,pad=0.3'
        ),
        verticalalignment='bottom' if y > 20 else 'top',
        zorder=zorder + 1
    )


def draw_single_model(ax: plt.Axes, img: np.ndarray, case: dict, model_name: str):
    ax.imshow(img)
    ax.axis('off')

    gt_cat = case['category_name']
    pred_bbox = case.get(f'{model_name}_bbox')
    pred_iou = case.get(f'{model_name}_iou', 0)
    pred_score = case.get(f'{model_name}_score', 0)
    pred_category = case.get(f'{model_name}_pred_category')
    localized = case.get(f'{model_name}_localized', False)
    classified_correct = case.get(f'{model_name}_classified_correct', False)

    if pred_bbox is not None:
        if classified_correct:
            color = COLORS['correct']
            status = "Correct"
        elif localized:
            color = COLORS['incorrect']
            status = f"Wrong: {pred_category}"
        else:
            color = COLORS['missed']
            status = f"Low IoU: {pred_category}"

        if not classified_correct:
            pred_text = pred_category if pred_category else "None"
            label = f"GT: {gt_cat}\nPred: {pred_text}"
        else:
            label = pred_category if pred_category else "?"

        draw_bbox(
            ax, pred_bbox, label, color,
            score=pred_score,
            linewidth=3.0,
            zorder=2
        )

        title = f"{MODEL_NAMES.get(model_name, model_name)}\nIoU: {pred_iou:.3f} | {status}"
    else:
        title = f"{MODEL_NAMES.get(model_name, model_name)}\nNo Detection"

    if classified_correct:
        title_color = COLORS['correct']
    elif localized:
        title_color = COLORS['incorrect']
    else:
        title_color = COLORS['missed']

    ax.set_title(title, fontweight='bold', fontsize=10, color=title_color)


def create_comparison_figure(img: np.ndarray, case: dict, output_path: Path, dpi: int, case_type: str):
    img_h, img_w = img.shape[:2]
    aspect_ratio = img_w / img_h

    num_models = len(MODEL_ORDER)
    fig_width = 11
    single_width = fig_width / num_models
    fig_height = single_width / aspect_ratio + 1.8

    fig, axes = plt.subplots(1, num_models, figsize=(fig_width, fig_height))
    if num_models == 1:
        axes = [axes]

    for ax, model_name in zip(axes, MODEL_ORDER):
        draw_single_model(ax, img, case, model_name)

    cat_name = case['category_name']
    size_cat = case['size_category']

    if case_type == 'advantage':
        advantage_type = case.get('advantage_type', 'N/A').replace('_', ' ').title()
        suptitle = f"GT: {cat_name} (Rare, {size_cat.capitalize()}) | DeCLIP Advantage: {advantage_type}"
    else:
        declip_iou = case.get('declip_iou', 0)
        declip_pred_cat = case.get('declip_pred_category', 'N/A')
        suptitle = f"GT: {cat_name} (Rare, {size_cat.capitalize()}) | DeCLIP: {declip_pred_cat} (IoU: {declip_iou:.2f})"

    fig.suptitle(suptitle, fontsize=12, fontweight='bold', y=0.98)

    legend_elements = [
        patches.Patch(facecolor=COLORS['correct'], edgecolor='none', label='Correct Classification'),
        patches.Patch(facecolor=COLORS['incorrect'], edgecolor='none', label='Wrong Classification'),
        patches.Patch(facecolor=COLORS['missed'], edgecolor='none', label='Not Localized (IoU<0.5)')
    ]
    fig.legend(
        handles=legend_elements,
        loc='lower center',
        ncol=3,
        frameon=True,
        fontsize=10,
        bbox_to_anchor=(0.5, 0.02),
        markerscale=1.5
    )

    plt.tight_layout(rect=[0, 0.08, 1, 0.95])
    plt.savefig(output_path, dpi=dpi, bbox_inches='tight', facecolor='white', edgecolor='none')
    plt.close()


def create_single_model_figure(img: np.ndarray, case: dict, model_name: str, output_path: Path, dpi: int):
    img_h, img_w = img.shape[:2]
    aspect_ratio = img_w / img_h

    fig_width = 8
    fig_height = fig_width / aspect_ratio + 1.0
    fig, ax = plt.subplots(1, 1, figsize=(fig_width, fig_height))

    draw_single_model(ax, img, case, model_name)

    legend_elements = [
        patches.Patch(facecolor=COLORS['correct'], edgecolor='none', label='Correct'),
        patches.Patch(facecolor=COLORS['incorrect'], edgecolor='none', label='Wrong')
    ]
    ax.legend(
        handles=legend_elements,
        loc='upper right',
        fontsize=10,
        framealpha=0.9,
        markerscale=1.5
    )

    plt.tight_layout()
    plt.savefig(output_path, dpi=dpi, bbox_inches='tight', facecolor='white', edgecolor='none')
    plt.close()


def generate_index_html(cases: List[dict], output_dir: Path, case_type: str):
    if case_type == 'advantage':
        title = "DeCLIP Classification Advantage Cases (Rare)"
        subtitle = "Rare categories where DeCLIP correctly classifies and CLIPSelf fails"
        description = "Ranked by advantage score."
    else:
        title = "DeCLIP Unsolved Rare Cases"
        subtitle = "Rare categories where DeCLIP fails to classify correctly"
        description = "Ranked by IoU (descending)."

    html_content = f"""<!DOCTYPE html>
<html>
<head>
    <title>OV-LVIS Analysis - {title}</title>
    <style>
        body {{ font-family: Arial, sans-serif; margin: 20px; background: #f5f5f5; }}
        h1 {{ color: #2C3E50; }}
        .case-grid {{ display: grid; grid-template-columns: repeat(auto-fill, minmax(450px, 1fr)); gap: 20px; }}
        .case-card {{ background: white; border-radius: 8px; padding: 15px; box-shadow: 0 2px 5px rgba(0,0,0,0.1); }}
        .case-card img {{ width: 100%; border-radius: 4px; }}
        .case-info {{ margin-top: 10px; font-size: 14px; line-height: 1.6; }}
        .tag {{ display: inline-block; padding: 2px 8px; border-radius: 4px; margin-right: 5px; font-size: 12px; }}
        .tag-rare {{ background: #8E44AD; color: white; }}
        .tag-small {{ background: #3498DB; color: white; }}
        .correct {{ color: #27AE60; font-weight: bold; }}
        .incorrect {{ color: #E74C3C; font-weight: bold; }}
        .model-result {{ margin: 5px 0; padding: 5px; background: #f9f9f9; border-radius: 4px; }}
    </style>
</head>
<body>
    <h1>Open-Vocabulary LVIS Analysis</h1>
    <h2>{subtitle}</h2>
    <p>{description}</p>

    <div class="case-grid">
"""

    for idx, case in enumerate(cases):
        cat_name = case['category_name']
        size_cat = case['size_category']
        img_id = case['image_id']

        case_folder = f"rank{idx+1:02d}_{cat_name}_{size_cat}_img{img_id}"

        def get_cls_html(prefix, model_name):
            pred_cat = case.get(f'{prefix}_pred_category', 'N/A')
            correct = case.get(f'{prefix}_classified_correct', False)
            iou = case.get(f'{prefix}_iou', 0)
            if correct:
                return f'<span class="correct">{model_name}: {pred_cat} ✓ (IoU: {iou:.2f})</span>'
            if pred_cat:
                return f'<span class="incorrect">{model_name}: {pred_cat} ✗ (IoU: {iou:.2f})</span>'
            return f'<span class="incorrect">{model_name}: No detection</span>'

        declip_html = get_cls_html('declip', 'DeCLIP')
        clipself_html = get_cls_html('clipself', 'CLIPSelf')

        if case_type == 'advantage':
            advantage_type = case.get('advantage_type', 'N/A').replace('_', ' ').title()
            status_html = f'<strong>Advantage: {advantage_type}</strong>'
        else:
            status_html = '<strong class="incorrect">Unsolved</strong>'

        html_content += f"""
        <div class="case-card">
            <a href="{case_folder}/comparison.png" target="_blank">
                <img src="{case_folder}/comparison.png" alt="Case {idx+1}">
            </a>
            <div class="case-info">
                <strong>Rank #{idx+1}</strong> - GT: {cat_name}<br>
                <span class="tag tag-rare">Rare</span>
                <span class="tag tag-{size_cat}">{size_cat.capitalize()}</span>
                {status_html}<br>
                <div class="model-result">{declip_html}</div>
                <div class="model-result">{clipself_html}</div>
            </div>
        </div>
"""

    html_content += """
    </div>
</body>
</html>
"""

    with open(output_dir / 'index.html', 'w') as f:
        f.write(html_content)

    print(f"  Index page saved to: {output_dir / 'index.html'}")


def generate_summary_figure(cases: List[dict], output_dir: Path, dpi: int, case_type: str):
    by_category = defaultdict(int)
    for case in cases:
        by_category[case['category_name']] += 1

    sorted_cats = sorted(by_category.items(), key=lambda x: x[1], reverse=True)[:10]
    cat_names = [c[0] for c in sorted_cats]
    cat_counts = [c[1] for c in sorted_cats]

    if case_type == 'advantage':
        fig, axes = plt.subplots(1, 2, figsize=(12, 5))

        axes[0].barh(cat_names[::-1], cat_counts[::-1], color='#3498DB', edgecolor='white', linewidth=1.5)
        axes[0].set_xlabel('Number of Cases', fontweight='bold')
        axes[0].set_title('DeCLIP Advantage by Rare Category', fontweight='bold')

        by_adv = defaultdict(int)
        for case in cases:
            by_adv[case.get('advantage_type', 'unknown')] += 1

        adv_types = ['unique_correct_classification', 'better_iou']
        adv_labels = ['Only DeCLIP\nCorrect', 'Better\nIoU']
        adv_counts = [by_adv[t] for t in adv_types]
        colors_adv = ['#27AE60', '#3498DB']

        axes[1].bar(adv_labels, adv_counts, color=colors_adv, edgecolor='white', linewidth=1.5)
        axes[1].set_xlabel('Advantage Type', fontweight='bold')
        axes[1].set_ylabel('Number of Cases', fontweight='bold')
        axes[1].set_title('DeCLIP Advantage Types', fontweight='bold')
        for i, c in enumerate(adv_counts):
            axes[1].text(i, c + 0.5, str(c), ha='center', fontweight='bold')
    else:
        fig, axes = plt.subplots(1, 2, figsize=(12, 5))

        axes[0].barh(cat_names[::-1], cat_counts[::-1], color='#E74C3C', edgecolor='white', linewidth=1.5)
        axes[0].set_xlabel('Number of Cases', fontweight='bold')
        axes[0].set_title('DeCLIP Unsolved Cases by Rare Category', fontweight='bold')

        confusion = defaultdict(int)
        for case in cases:
            gt_cat = case['category_name']
            pred_cat = case.get('declip_pred_category', 'N/A')
            if pred_cat and pred_cat != gt_cat:
                confusion[(gt_cat, pred_cat)] += 1

        sorted_conf = sorted(confusion.items(), key=lambda x: x[1], reverse=True)[:10]
        conf_labels = [f"{c[0][0]}→{c[0][1]}" for c in sorted_conf]
        conf_counts = [c[1] for c in sorted_conf]

        if conf_labels:
            axes[1].barh(conf_labels[::-1], conf_counts[::-1], color='#9B59B6', edgecolor='white', linewidth=1.5)
            axes[1].set_xlabel('Number of Cases', fontweight='bold')
            axes[1].set_title('Top Misclassification Patterns (GT→Pred)', fontweight='bold')
        else:
            axes[1].text(0.5, 0.5, 'No misclassification data', ha='center', va='center', transform=axes[1].transAxes)
            axes[1].set_title('Misclassification Patterns', fontweight='bold')

    plt.tight_layout()
    plt.savefig(output_dir / 'summary_statistics.png', dpi=dpi, bbox_inches='tight', facecolor='white')
    plt.close()

    print(f"  Summary figure saved to: {output_dir / 'summary_statistics.png'}")


def main():
    args = parse_args()

    print("=" * 70)
    print("OV-LVIS Visualization Generator (Rare Categories)")
    print("=" * 70)

    with open(args.cases, 'r') as f:
        cases = json.load(f)
    print(f"  Total cases to visualize: {len(cases)}")

    img_id_to_file = load_image_map(args.ann_file)
    output_dir = Path(args.output)
    output_dir.mkdir(parents=True, exist_ok=True)

    for idx, case in enumerate(cases):
        img_id = case['image_id']
        cat_name = case['category_name']
        size_cat = case['size_category']

        case_dir = output_dir / f"rank{idx+1:02d}_{cat_name}_{size_cat}_img{img_id}"
        case_dir.mkdir(parents=True, exist_ok=True)

        file_name = img_id_to_file.get(img_id)
        if not file_name:
            print(f"  [!] Image id not found: {img_id}")
            continue

        img_path = Path(args.img_dir) / file_name
        if not img_path.exists():
            print(f"  [!] Image not found: {img_path}")
            continue

        img = np.array(Image.open(img_path).convert('RGB'))

        create_single_model_figure(img, case, 'declip', case_dir / 'declip.png', args.dpi)
        create_single_model_figure(img, case, 'clipself', case_dir / 'clipself.png', args.dpi)

        create_comparison_figure(img, case, case_dir / 'comparison.png', args.dpi, args.case_type)

        with open(case_dir / 'info.json', 'w') as f:
            json.dump(case, f, indent=2)

        print(f"  [{idx+1}/{len(cases)}] {case_dir.name}")

    print("\nGenerating index page...")
    generate_index_html(cases, output_dir, args.case_type)

    print("Generating summary figure...")
    generate_summary_figure(cases, output_dir, args.dpi, args.case_type)

    print(f"\nVisualization complete: {output_dir}")


if __name__ == '__main__':
    main()