Upload README.md

README.md

@@ -1,143 +1,293 @@
 ---
+pretty_name: CDBench
 license: cc-by-nc-sa-4.0
+language:
+- en
 task_categories:
+- visual-question-answering
 - question-answering
 size_categories:
 - 10K<n<100K
-dataset_info:
-# - config_name: viewer
-#   features:
-#   - name: question
-#     dtype: string
-#   - name: options
-#     dtype: string
-#   - name: answer
-#     dtype: string
-#   - name: target_image
-#     dtype: image
-#   - name: reference_image
-#     dtype: image
-#   - name: mask
-#     dtype: image
+tags:
+- change-detection
+- multimodal
+- vision-language
+- benchmark
+- remote-sensing
+- industrial-inspection
+- image
+- text
 configs:
 - config_name: viewer
-  data_files: "questions.csv"
+  data_files:
+  - split: train
+    path: questions.csv
 ---
 
 # CDBench: A Comprehensive Multimodal Dataset and Evaluation Benchmark for General Change Detection
 
-[![Demo](https://img.shields.io/badge/🤗-Demo-green)](http://demo.mm-cd.org:8880) 
+**16,086 image pairs · 73,122 multiple-choice QA items · 7 tasks · 3 domains**
 
-_Demo response may be slow due to limited resources. We appreciate your understanding._
+[Project website](http://mm-cd.org) | [Dataset files](https://huggingface.co/datasets/MM-CD/CDBench/tree/main) | [Questions and feedback](https://huggingface.co/datasets/MM-CD/CDBench/discussions)
 
-## Introduction
+CDBench evaluates how multimodal large language models (MLLMs) understand images and reason about changes across **remote sensing, industrial inspection, and commodity products**. It connects image content recognition with change discrimination, regional localization, semantic classification, description, and plausible cause inference through a shared multiple-choice question (MCQ) format.
 
-General change detection aims to identify and interpret meaningful differences observed in scenes or objects across different states, playing a critical role in domains such as remote sensing and industrial inspection. While Multimodal Large Language Models (MLLMs) show promise, their capabilities in structured general change detection remain underexplored.
+The accompanying **ChangeAgent** framework combines an MLLM with expert visual guidance and retrieval-augmented generation. In the reported evaluation, it achieves **78.25%** average accuracy across Q1–Q7, compared with **71.19%** for Gemini 1.5 Pro and **70.68%** for GPT-4o.
 
-This project introduces **CDBench**, the first comprehensive benchmark for evaluating MLLMs' capabilities in multimodal general change detection across diverse domains. Our benchmark unifies diverse datasets and defines seven structured tasks: two image analysis tasks (Image Content Classification, Image Content Description) and five change analysis tasks (Change Discrimination, Change Localization, Semantic Change Classification/Detection, Change Description, and Change Reasoning).
+## Dataset at a glance
 
-We also propose the **ChangeAgent** framework, which enhances MLLM cores through retrieval-augmented generation and expert visual guidance, achieving a significantly higher average accuracy of 77.10% on CDBench, compared to 70-71% for leading baseline MLLMs.
+The counts below are computed directly from the released `questions.csv`. An image pair is a unique `(target_image, reference_image)` tuple; a QA item is one CSV row.
+
+| Domain | Source dataset | Image pairs | QA items | Task coverage |
+|---|---|---:|---:|---|
+| Remote sensing | CDD | 2,343 | 9,468 | Q1–Q4, Q6–Q7 |
+| Remote sensing | LEVIR-CD | 3,093 | 12,444 | Q1–Q4, Q6–Q7 |
+| Remote sensing | SYSU-CD | 2,316 | 9,409 | Q1–Q4, Q6–Q7 |
+| Industrial inspection | MVTec-AD | 1,685 | 9,442 | Q1–Q7 |
+| Industrial inspection | MVTec-LOCO | 1,557 | 8,016 | Q1–Q7 |
+| Industrial inspection | VisA | 2,132 | 10,189 | Q1–Q7 |
+| Commodity products | GoodsAD | 2,960 | 14,154 | Q1–Q7 |
+| **Total** | **7 source datasets** | **16,086** | **73,122** | **7 tasks overall** |
+
+Not every image pair has all seven questions. The released remote-sensing subsets contain no Q5 items, and coverage varies across tasks and samples. The totals describe the released annotations, not the sizes of the original source datasets.
+
+## Seven evaluation tasks
+
+| ID | Task | What the model selects | QA items |
+|---|---|---|---:|
+| Q1 | Image Content Classification | The primary scene type or dominant content category of one image. | 16,075 |
+| Q2 | Image Content Description | The description that best matches objects, attributes, composition, and spatial relations in one image. | 8,995 |
+| Q3 | Change Discrimination | Whether a meaningful change is present relative to the reference image. | 16,038 |
+| Q4 | Change Localization | A regional or semantic location where the change occurs. | 8,991 |
+| Q5 | Semantic Change Classification | The category of the observed change or defect. | 5,034 |
+| Q6 | Change Description | The description that best summarizes the observed differences. | 8,995 |
+| Q7 | Change Cause Inference | A plausible cause or underlying reason for the change. | 8,994 |
+
+**Q1–Q2 assess single-image understanding; Q3–Q7 assess paired-image change analysis. All seven tasks use MCQ answer selection**, including Q2 and Q6. Q4 evaluates location-related answers, not pixel-level segmentation. Q7 evaluates plausible interpretation from available evidence, not experimentally established causality.
 
 <div align="center">
-  <img src="assets/overview.png" width="80%">
-  <br>
-  <em>Figure 1: Overview of CDBench benchmark tasks and ChangeAgent architecture</em>
+  <img src="assets/cdbench-examples-v26.png" alt="Six CDBench examples covering image analysis and paired-image change analysis" width="100%">
+  <p><em>Representative MCQs. Shaded options mark annotated answers. Panels (a)–(b) use single images; in (c)–(f), the reference/earlier image appears above the target/later image.</em></p>
 </div>
 
-## What's New
+## Dataset construction
 
-- **First Multimodal Change Detection Benchmark**: Introduced the first comprehensive benchmark specifically designed for evaluating MLLMs on general change detection tasks
-- **Seven-Task Evaluation Framework**: Designed seven structured tasks covering image analysis and change analysis, forming a large-scale evaluation set with 70,000+ question-answer pairs
-- **Hybrid Cross-Generation Strategy**: Employed LLM-driven content generation, cross-model optimization strategies, and dual-expert human validation to ensure evaluation quality and fairness
-- **ChangeAgent Innovative Architecture**: Proposed a hybrid framework combining expert visual guidance and retrieval-augmented generation, significantly improving change detection performance
+1. **Integrate visual sources.** The benchmark combines seven datasets across three domains. Where a source lacks paired references, nearest-neighbor retrieval selects relevant normal examples. For remote-sensing images with limited prior information, CLIP-based scene classification supplies additional context.
+2. **Draft and refine questions across models.** Available images, masks, labels, descriptions, and domain knowledge support question construction. GPT-4o drafts questions that Claude reviews and refines, and the models also exchange these roles.
+3. **Verify with two experts.** Experts inspect and correct QA items for question relevance, answer validity, and ambiguity through a collaborative annotation platform.
 
-## Dataset
+This procedure is designed to reduce generator-specific bias and ambiguous answers. Masks and labels used to ground annotation are **not automatically inputs to the evaluated MLLMs**.
 
-### Dataset Overview
+<div align="center">
+  <img src="assets/cdbench-construction-v26.png" alt="Shared annotation context, reciprocal GPT-4o and Claude refinement, dual-expert verification, and seven MCQ tasks" width="100%">
+  <p><em>CDBench construction and task organization. The verification cards illustrate a correct answer and an intentionally incorrect alternative; they are examples of the checks, not historical rejection records.</em></p>
+</div>
 
-CDBench integrates diverse datasets from multiple domains, including remote sensing, industrial inspection, and commodity product change/anomaly detection, totaling over 15,000 image pairs. The dataset covers:
+## Repository files and availability
+
+The current release contains `questions.csv`, six image archives, and documentation images in `assets/`:
+
+```text
+questions.csv
+CDD.zip
+LEVIR-CD.zip
+SYSU-CD.zip
+MVTec-AD.zip
+MVTec-LOCO.zip
+VisA.zip
+assets/
+README.md
+```
+
+**GoodsAD availability:** `questions.csv` includes 14,154 GoodsAD QA items covering 2,960 image pairs, but this repository does not currently contain a `GoodsAD.zip` archive. Downloading the six available archives therefore does not provide all images referenced by the CSV. For the matching GoodsAD files, contact the maintainers through the [Community tab](https://huggingface.co/datasets/MM-CD/CDBench/discussions). If evaluating only available subsets, report that subset explicitly.
 
-- **Remote Sensing Data**: LEVIR-CD, SYSU-CD, CDD
-- **Industrial Inspection**: MVTec-AD, MVTec-LOCO, Visa  
-- **Commodity Inspection**: GoodsAD
+The Hugging Face viewer exposes the annotations as the `viewer` configuration with a split named **`train`**. This is the current file-loading configuration; it does **not** define an official training/evaluation partition. The reported benchmark evaluates models without CDBench-specific fine-tuning.
 
-To address the issue of some original data lacking paired reference samples, we employ nearest neighbor search to retrieve the most relevant reference samples from normal categories. For remote sensing data, we utilize CLIP models to perform scene classification based on approximately 50 predefined scene categories, providing richer contextual information for subsequent analysis.
+### Annotation schema
 
-### Dataset Statistics
+| Field | Type | Meaning |
+|---|---|---|
+| `target_image` | string | Relative path to the target image. |
+| `question_num` | integer | Task identifier, from 1 to 7. |
+| `question` | string | Question stem in English. |
+| `options` | string | Lettered answer choices, usually separated by newlines. |
+| `answer` | string | Correct option letter: `A`, `B`, `C`, or `D`. |
+| `reference_image` | string | Relative path to the reference image. |
+| `mask` | string or empty | Relative path to an available change/defect mask; some entries are empty. |
 
-| Dataset Category | Image Pairs | Tasks | QA Pairs |
-|------------------|-------------|-------|----------|
-| Remote Sensing   | 7,000+      | 7     | 30,000+  |
-| Industrial Inspection | 5,000+ | 7     | 30,000+  |
-| Commodity Inspection | 2,000+ | 7     | 10,000+  |
-| **Total**        | **14,000+** | **7** | **70,000+** |
+Image fields contain **paths**, not embedded image bytes. The CSV contains quoted multiline option strings, so use a CSV parser rather than splitting the file by line. An empty mask field means no mask path is supplied for that row; it should not be treated as a negative-change label.
 
-### Data Examples
+One released Q4 record is shown below. Newlines in `options` are escaped here for readability.
 
-<div align="center">
-  <img src="assets/dataset_examples.png" width="80%">
-  <br>
-  <em>Figure 2: Examples of seven tasks in the CDBench dataset</em>
-</div>
+```json
+{
+  "target_image": "CDD/B/10_154.png",
+  "question_num": 4,
+  "question": "Where have the most significant changes occurred between Time Points A and B?",
+  "options": "A. In the agricultural regions;\nB. Near large water bodies.;\nC. In the downtown urban area;\nD. Along the roadway infrastructure",
+  "answer": "D",
+  "reference_image": "CDD/A/10_154.png",
+  "mask": "CDD/label/10_154.png"
+}
+```
 
-## Seven Core Tasks
+## Quick start
 
-1. **Q1: Image Content Classification** - Identify the primary scene type or dominant content category of a given image
-2. **Q2: Image Content Description** - Generate comprehensive textual descriptions of images, detailing fine-grained features and spatial layouts
-3. **Q3: Change Discrimination** - Determine whether significant changes have occurred between two registered images
-4. **Q4: Change Localization** - Identify and describe the specific pixel or regional locations where changes occur
-5. **Q5: Semantic Change Classification** - Classify the nature or semantic category of detected changes
-6. **Q6: Change Description** - Provide concise textual summaries of identified change events
-7. **Q7: Change Reasoning** - Infer plausible causes or root reasons for detected changes
+### Load the annotations
 
-## Demo
+```bash
+pip install datasets huggingface_hub pillow
+```
 
-We provide an online demonstration system accessible through the following links:
+```python
+from datasets import load_dataset
 
-- 🌐 [Official Project Website](http://mm-cd.org)
-- 📊 [Interactive Evaluation Platform](http://demo.mm-cd.org:8880) Demo response may be slow due to limited resources. We appreciate your understanding.
+qa = load_dataset("MM-CD/CDBench", name="viewer", split="train")
+print(qa.num_rows)  # 73122 in this release
+print(qa[0]["question"])
+print(qa[0]["options"])
 
-<div align="center">
-  <img src="assets/demo_screenshot.png" width="80%">
-  <br>
-  <em>Figure 3: Screenshot of CDBench online evaluation platform</em>
-</div>
+# Example: select all regional-localization questions.
+q4 = qa.filter(lambda row: row["question_num"] == 4)
+```
+
+Loading this CSV configuration retrieves the annotations; it does not download or decode the image archives.
+
+### Download image archives
+
+Download only the source datasets you need. For example:
+
+```python
+from huggingface_hub import hf_hub_download
+
+archive = hf_hub_download(
+    repo_id="MM-CD/CDBench",
+    repo_type="dataset",
+    filename="CDD.zip",
+    local_dir="./CDBench",
+)
+print(archive)
+```
+
+Repeat with another filename from the repository listing as needed. Each archive is several gigabytes; the six archives total approximately 25.24 GB before extraction.
+
+**Arrange extracted files to match the CSV paths.** Archive directory layouts differ: the remote-sensing archives start with an `mmcd_set/` wrapper, while the industrial archives start with object-category folders. Inspect the extracted tree and place each source under its dataset-name directory, avoiding extra wrapper levels. For example, the resulting image root should contain:
+
+```text
+images/
+├── CDD/
+│   ├── A/10_154.png
+│   ├── B/10_154.png
+│   └── label/10_154.png
+├── LEVIR-CD/
+├── SYSU-CD/
+├── MVTec-AD/
+├── MVTec-LOCO/
+└── VisA/
+```
+
+After arranging the files, open a pair using the paths in its record:
+
+```python
+from pathlib import Path
+from PIL import Image
+
+image_root = Path("./CDBench/images")
+row = qa[0]
 
-## Results
+with Image.open(image_root / row["reference_image"]) as image:
+    reference = image.convert("RGB")
+with Image.open(image_root / row["target_image"]) as image:
+    target = image.convert("RGB")
 
-### Performance Comparison
+print(reference.size, target.size)
+```
 
-| Model | Image Classification (↑) | Image Description (↑) | Change Discrimination (↑) | Change Localization (↑) | Change Classification (↑) | Change Description (↑) | Change Reasoning (↑) | Average (↑) |
-|-------|-------------------------|---------------------|-------------------------|------------------------|-------------------------|----------------------|-------------------|-------------|
+The official [Hugging Face loading guide](https://huggingface.co/docs/datasets/loading) and [Hub download guide](https://huggingface.co/docs/huggingface_hub/guides/download) provide additional loading and revision-pinning options.
+
+## Evaluation protocol
+
+- **Metric:** top-1 MCQ accuracy for each of Q1–Q7. `Avg.` is the unweighted arithmetic mean of the seven task accuracies, not a pooled average over all QA rows.
+- **Cascading scoring:** an incorrect foundational answer causes subsequent dependent answers for the same sample to be counted as incorrect. Later-task scores therefore reflect both the task itself and upstream judgments.
+- **Setting:** zero-shot prompting without task-specific fine-tuning on CDBench. Each evaluation is run twice, and the reported scores average the two runs.
+- **Inputs:** distinguish the single-image tasks from paired-image tasks. Keep answers and annotation-only information out of model prompts; report any additional masks, retrieved context, or expert guidance used by a system.
+
+For comparable results, report the evaluated subsets, model version, prompts, input context, per-task scores, and scoring protocol. Independent per-question accuracy without cascading is a different metric. The current repository provides data files; it does not contain a released scoring script or machine-readable task-dependency specification.
+
+### Reported model performance
+
+Accuracy (%). **Bold marks the best value in each column.** These are the reported benchmark runs, not evaluations of the latest versions of the named model families.
+
+| Model | Q1 | Q2 | Q3 | Q4 | Q5 | Q6 | Q7 | Avg. |
+|---|---:|---:|---:|---:|---:|---:|---:|---:|
 | Qwen-Max-Latest | 57.03 | 79.97 | 55.88 | 72.33 | 61.78 | 64.35 | 65.00 | 65.19 |
 | Qwen-plus-latest | 55.12 | 79.04 | 55.87 | 72.18 | 62.77 | 63.70 | 65.93 | 64.94 |
+| Qwen-turbo-1101 | 48.90 | 68.15 | 55.93 | 66.07 | 54.63 | 57.25 | 59.82 | 58.68 |
+| Grok-3 | 73.16 | 80.38 | 59.67 | 68.51 | 60.58 | 61.70 | 62.26 | 66.61 |
 | Claude-3-5-sonnet | 70.17 | 82.27 | 67.47 | 71.74 | 64.96 | 58.48 | 64.15 | 68.46 |
+| Claude-3-7-sonnet | 65.87 | 75.26 | **79.87** | 65.40 | 61.58 | 59.92 | 61.26 | 67.02 |
+| Gemini-1.5-flash | 48.21 | 70.15 | 57.30 | 63.51 | 60.00 | 55.81 | 56.15 | 58.73 |
 | Gemini-1.5-pro | 72.54 | 83.07 | 66.03 | 73.96 | 66.55 | 67.59 | 68.60 | 71.19 |
-| GPT-4o | 91.37 | 85.00 | 69.72 | 69.70 | 56.83 | 61.71 | 60.46 | 70.68 |
-| **ChangeAgent (Ours)** | **96.87** | **76.78** | **78.81** | **76.79** | **77.67** | **70.82** | **69.99** | **77.10** |
+| GPT-4o | 91.37 | **85.00** | 69.72 | 69.70 | 56.83 | 61.71 | 60.46 | 70.68 |
+| ChangeAgent | **96.87** | 76.78 | 78.81 | **76.79** | **77.67** | **70.82** | **69.99** | **78.25** |
 
-### ChangeAgent Architecture Advantages
+ChangeAgent improves the average by **7.06 percentage points** over Gemini 1.5 Pro and **7.57 points** over GPT-4o. It ranks first on Q1 and Q4–Q7; GPT-4o leads Q2, and Claude-3-7-sonnet leads Q3.
 
-ChangeAgent achieves performance improvements through three core modules:
+<div align="center">
+  <img src="assets/cdbench-performance-v26.png" alt="Model comparison using the same ten model rows and accuracies as the table" width="100%">
+  <p><em>Task-group means, overall means, and individual Q1–Q7 accuracies from the table above. Shading identifies ChangeAgent.</em></p>
+</div>
 
-1. **Expert Visual Change Localization**: Utilizes CLIP visual encoder and change decoder for precise change localization
-2. **Knowledge Retrieval and Enhancement**: Injects domain-specific prior knowledge through RAG module
-3. **Integrated Reasoning and Task Execution**: Combines multimodal context for complex semantic reasoning
+### ChangeAgent
 
-### Visualization Results
+ChangeAgent combines paired CLIP visual features, an aggregated change map, retrieved domain knowledge, and task inputs for MLLM reasoning. The change map supplies spatial evidence, while the original visual features retain broader scene semantics. Retrieval adds relevant priors, task definitions, and change examples.
 
 <div align="center">
-  <img src="assets/results_comparison.png" width="90%">
-  <br>
-  <em>Figure 4: Performance comparison between ChangeAgent and existing MLLMs on complex change detection tasks</em>
+  <img src="assets/changeagent-framework-v26.png" alt="ChangeAgent architecture with shared visual encoding, change decoding, retrieval augmentation, and MLLM reasoning" width="100%">
+  <p><em>ChangeAgent uses both visual guidance and retrieved knowledge for seven-task reasoning.</em></p>
 </div>
 
+Because it uses additional visual guidance and retrieval, ChangeAgent serves as a **hybrid reference**, rather than an input-matched MLLM baseline. The reported comparison does not isolate the contribution of each component; that requires ablation experiments.
+
+### Visual dependence and language bias
+
+GPT-4o was evaluated with and without images on the **same randomly sampled 1,000 image pairs (4,628 QA items)** under the cascading protocol.
+
+| Input | Avg. (Q1–Q7) | Q4 | Q5 | Q6 | Q7 |
+|---|---:|---:|---:|---:|---:|
+| With images | 70.68 | 69.70 | 56.83 | 61.71 | 60.46 |
+| Text-only | 32.93 | 26.39 | 15.74 | 22.74 | 21.88 |
+
+Removing images reduces the average by **37.75 percentage points**, showing a substantial contribution from visual evidence. This experiment does **not** establish the absence of residual language bias. Cascading also propagates upstream errors, so later-task scores should be interpreted in that context.
+
+## Scope and limitations
+
+CDBench measures structured visual comparison and answer selection across the included domains. It complements pixel-level change-detection benchmarks and free-form captioning evaluations; its accuracy should not be interpreted as segmentation quality or unrestricted generation quality. Retrieved normal references in anomaly datasets are not necessarily temporal observations of the same object, and the image pairs are not universally registered.
+
+The source datasets, task frequencies, and answer formats are heterogeneous. Report domain and task coverage when comparing systems. Cross-model refinement and expert verification aim to improve annotation quality, but do not guarantee that all ambiguity, language priors, or model-specific bias has been eliminated.
+
+## Project and demo
+
+- [Project website](http://mm-cd.org)
+- [Interactive demo](http://demo.mm-cd.org:8880) — availability and response times depend on the demo server.
+- [Community discussions](https://huggingface.co/datasets/MM-CD/CDBench/discussions) for data issues, missing files, and questions.
+
+## License and acknowledgements
+
+This repository is distributed with the **CC BY-NC-SA 4.0** license tag. Images originate from the seven source datasets listed above; consult their original licenses and attribution requirements as well.
+
+We thank the creators of LEVIR-CD, SYSU-CD, CDD, MVTec-AD, MVTec-LOCO, VisA, and GoodsAD; the experts who annotated and verified the QA items; Fudan University; Shanghai Ocean University; Shanghai Vision Medical Technology Co., Ltd.; and the open-source community supporting the underlying vision-language models and tools.
+
+## Citation
 
-## Acknowledgements
+To cite this dataset release:
 
-We thank the following projects and institutions for their support:
-- School of Computer Science and Artificial Intelligence, Fudan University
-- School of Information Technology, Shanghai Ocean University  
-- R&D Department, Shanghai Vision Medical Technology Co., Ltd.
-- All experts who participated in data annotation and validation
-- Open source community for providing foundational models like CLIP and LLaMA
+```bibtex
+@misc{chen2026cdbench,
+  title  = {{CDBench}: A Comprehensive Multimodal Dataset and Evaluation Benchmark for General Change Detection},
+  author = {Chen, Jintao and Qian, Rui and Peng, Bo and Cheng, Linjie and Li, Xiang and Chen, Tian and Ye, Jingyong and Yang, Yinhao and Chi, Mingmin and Dou, Dejing},
+  year   = {2026},
+  url    = {https://huggingface.co/datasets/MM-CD/CDBench},
+  note   = {Dataset and benchmark}
+}
+```