Iris314 commited on
Commit
853c181
·
verified ·
1 Parent(s): a9294cb

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +59 -0
README.md ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Model Card for Lego Brick Classification (Classical AutoML)
2
+
3
+ ## Model Details
4
+ - **Model type**: Classical AutoML ensemble (AutoGluon Tabular)
5
+ - **Framework**: [AutoGluon](https://auto.gluon.ai/stable/index.html)
6
+ - **Task**: Multiclass Classification
7
+ - **Classes**:
8
+ - Standard
9
+ - Flat
10
+ - Sloped
11
+
12
+ The model was trained using AutoGluon Tabular with automated feature engineering, model selection, and hyperparameter tuning under a fixed budget.
13
+
14
+ ## Intended Use
15
+ - **Direct use**: Classify LEGO bricks into `Standard`, `Flat`, or `Sloped` types based on their dimensions.
16
+ - **Educational use**: Demonstrate classical AutoML techniques on a small synthetic dataset.
17
+ - **Not intended for**: Industrial-scale LEGO piece classification or manufacturing quality control.
18
+
19
+ ## Dataset
20
+ - **Dataset name**: [aedupuga/lego-sizes](https://huggingface.co/datasets/aedupuga/lego-sizes)
21
+ - **Description**: Contains dimensions and stud count of LEGO bricks, with labels indicating piece type.
22
+ - **Splits**:
23
+ - Original: 30 measured samples
24
+ - Augmented: 300 synthetically generated samples
25
+ - **Features**:
26
+ - Max Length (cm)
27
+ - Max Height (cm)
28
+ - Width (cm)
29
+ - Studs
30
+ - Type (Standard/Flat/Sloped)
31
+
32
+ ## Training Procedure
33
+ - **Train/Val/Test split**: 80/10/10 (stratified) on augmented dataset
34
+ - **External validation**: Performed on original dataset
35
+ - **AutoML settings**:
36
+ - Preset: `best_quality`
37
+ - Metric: Accuracy
38
+ - Time budget: ~30 minutes
39
+
40
+ ## Evaluation Results
41
+ - **Augmented Test Set**:
42
+ - Accuracy: 0.97
43
+ - Weighted F1: 0.96
44
+ - **Original External Validation Set**:
45
+ - Accuracy: 0.93
46
+ - Weighted F1: 0.92
47
+
48
+ ## Limitations
49
+ - The dataset is very small (only 30 original samples).
50
+ - Synthetic augmentation may not capture full real-world variability.
51
+ - Model may not generalize beyond this toy dataset.
52
+
53
+ ## Recommendations
54
+ - Use primarily for **demonstration and coursework**.
55
+ - Do not deploy in real-world LEGO production environments.
56
+
57
+ ## Model Card Contact
58
+ Author: *Xinxuan Tang (Carnegie Mellon University)* – xinxuant@andrew.cmu.edu
59
+ Dataset curator: Anuhya Edupuganti (CMU) – aedupuga@andrew.cmu.edu