File size: 2,860 Bytes
1dadeca
dbf3cf9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1dadeca
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
dbf3cf9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: cc-by-4.0
task_categories:
- image-to-image
language:
- en
pretty_name: CellOPC
size_categories:
- 100K<n<1M
tags:
- optical-proximity-correction
- inverse-lithography
- mask-optimization
- vlsi
- eda
- lithography
dataset_info:
  features:
  - name: image
    dtype: image
  - name: conditioning_image
    dtype: image
  - name: mask_type
    dtype: string
  - name: context
    dtype: int32
  - name: source_dataset
    dtype: string
  splits:
  - name: train
    num_bytes: 2488911027
    num_examples: 451912
  - name: validation
    num_bytes: 626298112
    num_examples: 112974
  - name: test
    num_bytes: 621060
    num_examples: 80
  download_size: 4285592645
  dataset_size: 3115830199
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
---

# CellOPC

CellOPC is a large-scale benchmark dataset for cell- and context-aware mask optimization. It is designed to support deep learning research for optical proximity correction (OPC) and inverse lithography technique (ILT) mask generation.

## Dataset Description

CellOPC is constructed from real integrated circuit layouts at the 45 nm technology node. Each sample is clipped around a standard-cell placement instance to preserve cell-level hierarchy and surrounding layout context. The dataset provides paired input layout/target images and optimized mask images for learning cell-wise mask generation under different context sizes.

The dataset contains both model-based OPC and ILT mask types. It is intended to evaluate how standard-cell identity, neighboring geometries, and input context size affect mask prediction and lithography-aware printability.

## Dataset Structure

Each sample contains the following fields:

- `conditioning_image`: input target/layout image.
- `image`: ground-truth optimized mask image.
- `mask_type`: mask generation type, such as `opc` or `ilt`.
- `context`: context size used when clipping the input layout.
- `source_dataset`: source subset name, such as `cellopc_opc_16`.

The dataset contains three splits:

| Split | Number of Examples |
|---|---:|
| train | 451,912 |
| validation | 112,974 |
| test | 80 |

## Intended Use

CellOPC is intended for:

- training image-to-image mask generation models;
- benchmarking deep learning methods for OPC and ILT;
- studying the impact of context size on mask prediction;
- evaluating cell-aware and context-aware mask optimization.

## Loading the Dataset

```python
from datasets import load_dataset

dataset = load_dataset("ChristyHu/CellOPC")

train_set = dataset["train"]
val_set = dataset["validation"]
test_set = dataset["test"]

sample = train_set[0]
layout = sample["conditioning_image"]
mask = sample["image"]
mask_type = sample["mask_type"]
context = sample["context"]