File size: 3,799 Bytes
4f9c5fa
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
# 3D Slicer Medical Imaging GUI Benchmark Dataset (CSV Format)

## Dataset Description

This dataset contains **315 end-to-end GUI automation tasks** for 3D Slicer medical imaging software, focusing on MRI brain analysis workflows.

### Dataset Summary

- **Total Tasks**: 315
- **Total Images**: 100 unique screenshots (file paths only)
- **Application**: 3D Slicer (medical imaging software)
- **Domain**: Medical imaging, MRI brain analysis
- **Format**: CSV with file paths (ultra memory-efficient)

### Supported Tasks

- GUI automation
- Medical imaging workflows
- Visual grounding
- Action prediction
- Task planning

## Dataset Structure

The dataset is provided as a CSV file with the following columns:

- `serial_number`: Task number (1-315)
- `task_id`: Unique identifier (e.g., "3dslicer_endtoend_001")
- `task`: Natural language task description
- `image_sequence`: Screenshot sequence (→ separated)
- `json_data`: Complete task data in JSON format
- `num_steps`: Number of steps in the trajectory
- `num_images`: Number of images for this task
- `image_paths`: Pipe-separated file paths to images
- `images_dir`: Base directory for images

### JSON Data Structure

The `json_data` field contains:

```json
{
  "id": "3dslicer_endtoend_001",
  "initial_state": {
    "application": "3D Slicer",
    "display_resolution": [1920, 1080],
    "loaded_image": "Import_Akash_Data.png"
  },
  "instruction": "Task description...",
  "trajectory": [
    {
      "step": 1,
      "action": "CLICK",
      "target": "Load Data (Akash)",
      "screenshot": "Import_Akash_Data.png",
      "note": "Step 1: Interacting with UI elements",
      "bbox": [1054, 0, 1089, 35]
    }
  ],
  "outputs": {
    "final_file": "task_1_output.png",
    "verification": {...},
    "success": true
  }
}
```

### Action Types

- **CLICK**: Button clicks, menu selections (71.1%)
- **SEGMENT**: Drawing ROIs, measurements (15.9%)
- **COMPLETE**: Task completion (5.8%)
- **TEXT**: Text input (3.2%)
- **ZOOM**: Zoom operations (2.0%)
- **SCROLL**: Navigation (2.0%)

## Usage

```python
import pandas as pd
import json
from PIL import Image
import os

# Load CSV dataset
df = pd.read_csv("3dslicer_benchmark.csv")

# Access a task
task = df.iloc[0]
print(f"Task: {task['task']}")
print(f"Steps: {task['num_steps']}")

# Parse JSON data
task_json = json.loads(task['json_data'])
print(f"Trajectory: {len(task_json['trajectory'])} steps")

# Load images on-demand
image_paths = task['image_paths'].split('|')
for i, img_path in enumerate(image_paths):
    if os.path.exists(img_path):
        img = Image.open(img_path)
        print(f"Image {i+1}: {img.size}")
```

## Memory Efficiency

This CSV-based approach provides:

-**Ultra-low memory usage** - no images loaded into memory
-**Fast loading** - CSV loads in seconds
-**Flexible access** - load images only when needed
-**Easy sharing** - single CSV file
-**Scalable** - works with any number of images

## Dataset Creation

This dataset was created using:
- Manual annotation of 3D Slicer workflows
- Automated bounding box extraction (red/orange/yellow highlights)
- Robust action inference with strict guardrails
- Ultra memory-efficient CSV processing

### Quality Assurance

- ✅ 100% consistent actions for same UI elements
- ✅ 100% consistent bounding boxes for same screenshots
- ✅ Only CLICK actions have bounding boxes
- ✅ All bounding boxes extracted from images
- ✅ Strict guardrails prevent inconsistencies
- ✅ Ultra memory-efficient processing

## Citation

```bibtex
@dataset{3dslicer_benchmark_2024,
  title={3D Slicer Medical Imaging GUI Benchmark Dataset},
  author={Rishu Kumar},
  year={2024},
  url={https://huggingface.co/datasets/rishuKumar404/MedUI_3DSlicer_CSV}
}
```

## License

MIT License