hoololi commited on
Commit
4921bca
·
verified ·
1 Parent(s): 490fffe

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +75 -59
README.md CHANGED
@@ -1,59 +1,75 @@
1
- ---
2
- dataset_info:
3
- features:
4
- - name: session_id
5
- dtype: string
6
- - name: question_id
7
- dtype: string
8
- - name: timestamp
9
- dtype: string
10
- - name: operand_a
11
- dtype: int64
12
- - name: operand_b
13
- dtype: int64
14
- - name: operation
15
- dtype: string
16
- - name: correct_answer
17
- dtype: int64
18
- - name: difficulty
19
- dtype: string
20
- - name: ocr_prediction
21
- dtype: string
22
- - name: ocr_parsed_number
23
- dtype: int64
24
- - name: is_correct
25
- dtype: bool
26
- - name: ocr_model_name
27
- dtype: string
28
- - name: ocr_processing_time
29
- dtype: float64
30
- - name: ocr_confidence
31
- dtype: float64
32
- - name: session_duration
33
- dtype: int64
34
- - name: session_total_questions
35
- dtype: int64
36
- - name: app_version
37
- dtype: string
38
- - name: hardware
39
- dtype: string
40
- - name: handwriting_image
41
- dtype: image
42
- - name: session_accuracy
43
- dtype: float64
44
- - name: session_total_ocr_time
45
- dtype: float64
46
- - name: session_avg_ocr_time
47
- dtype: float64
48
- splits:
49
- - name: train
50
- num_bytes: 31076.0
51
- num_examples: 10
52
- download_size: 49666
53
- dataset_size: 31076.0
54
- configs:
55
- - config_name: default
56
- data_files:
57
- - split: train
58
- path: data/train-*
59
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ task_categories:
4
+ - image-to-text
5
+ - optical-character-recognition
6
+ language:
7
+ - en
8
+ tags:
9
+ - handwritten-digits
10
+ - math-education
11
+ - ocr
12
+ size_categories:
13
+ - n<1K
14
+ ---
15
+
16
+ # CalcTrainer Dataset 🧮
17
+
18
+ Handwritten mathematical answers collected from the [CalcTrainer](https://huggingface.co/spaces/hoololi/CalcTrainer) interactive math training application.
19
+
20
+ ## Dataset Fields
21
+
22
+ ### Core Data
23
+ | Field | Type | Description |
24
+ |-------|------|-------------|
25
+ | `handwriting_image` | Image | Handwritten answer image (~100x100px) |
26
+ | `ocr_prediction` | string | Raw OCR output text |
27
+ | `ocr_parsed_number` | int32 | Cleaned numeric value from OCR |
28
+ | `is_correct` | bool | Whether OCR matches correct answer |
29
+
30
+ ### Mathematical Context
31
+ | Field | Type | Description |
32
+ |-------|------|-------------|
33
+ | `operand_a` | int32 | First number (e.g., 7 in "7 × 3") |
34
+ | `operand_b` | int32 | Second number (e.g., 3 in "7 × 3") |
35
+ | `operation` | string | Operation: `+`, `-`, `×`, `÷` |
36
+ | `correct_answer` | int32 | Expected correct answer |
37
+ | `difficulty` | string | `Facile` (Easy) or `Difficile` (Hard) |
38
+
39
+ ### OCR Metrics
40
+ | Field | Type | Description |
41
+ |-------|------|-------------|
42
+ | `ocr_model_name` | string | OCR model used (e.g., "microsoft/trocr-base-handwritten") |
43
+ | `ocr_processing_time` | float32 | Processing time in seconds |
44
+ | `hardware` | string | Hardware used for OCR |
45
+
46
+ ### Session Info
47
+ | Field | Type | Description |
48
+ |-------|------|-------------|
49
+ | `session_id` | string | Unique session identifier |
50
+ | `question_id` | string | Unique question identifier |
51
+ | `timestamp` | string | When the session was completed |
52
+ | `session_duration` | int32 | Session length (30 or 60 seconds) |
53
+ | `session_accuracy` | float32 | Overall session accuracy percentage |
54
+ | `session_avg_ocr_time` | float32 | Average OCR time per image in session |
55
+
56
+ ## Usage
57
+
58
+ ```python
59
+ from datasets import load_dataset
60
+
61
+ dataset = load_dataset("hoololi/CalcTrainer_dataset")
62
+ train_data = dataset["train"]
63
+
64
+ # Example: Access first item
65
+ item = train_data[0]
66
+ print(f"Math problem: {item['operand_a']} {item['operation']} {item['operand_b']} = {item['correct_answer']}")
67
+ print(f"OCR predicted: '{item['ocr_prediction']}' → {item['ocr_parsed_number']}")
68
+ print(f"Correct: {item['is_correct']}")
69
+ ```
70
+
71
+ ## Data Source
72
+
73
+ Real handwriting samples from users solving math problems in the CalcTrainer application. Users write answers on a digital canvas during timed math sessions.
74
+
75
+ **Generated from**: [CalcTrainer Interactive Math Training](https://huggingface.co/spaces/hoololi/CalcTrainer) 🧮