samanthajmichael commited on
Commit
ed5ae8a
·
verified ·
1 Parent(s): 5f2183d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +75 -3
README.md CHANGED
@@ -1,3 +1,75 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ metrics:
4
+ - accuracy
5
+ library_name: keras
6
+ ---
7
+ # Pottery Classification Model
8
+
9
+ ## Model Description
10
+ - **Task**: Multiclass Ceramic Pottery Classification
11
+ - **Architecture**: Multilayer Perceptron (MLP)
12
+ - **Input Features**: 95 archaeological ceramic features
13
+ - **Output Classes**: 9 distinct pottery types
14
+
15
+ ## Model Details
16
+ - **Performance**: 99.31% Test Accuracy
17
+ - **Test Loss**: 0.0376
18
+ - **Training Epochs**: 500
19
+ - **Batch Size**: 256
20
+
21
+ ## Dataset
22
+
23
+ **Name**: Ceramics: Temporal-Spatial Dataset
24
+ **Year**: 1988
25
+ **Source**: Digital Archaeological Record (tDAR)
26
+ **Identifier**:
27
+ - tDAR ID: 6039
28
+ - DOI: 10.6067/XCV8TD9WNB
29
+
30
+ **Description**:
31
+ Archaeological ceramic dataset containing 95 features across 9 distinct pottery types, collected to analyze spatial and temporal characteristics of ceramic artifacts.
32
+
33
+ **Features**:
34
+ - Total features: 95 after encoding
35
+ - Feature selection process: Detailed in companion [EDA Notebook](https://github.com/samanthajmichael/deep_learning/blob/main/deep_learning/Assignments/pottery_classifier/notebooks/01_EDA.ipynb)
36
+
37
+ | Feature | Description |
38
+ |---------|-------------|
39
+ | firing | Firing atmosphere |
40
+ | temper | Type of temper used |
41
+ | manipul | Surface manipulation |
42
+ | compact | Surface compaction |
43
+ | color | Paint colors used |
44
+ | pnttype | Paint type (organic, mineral, clay) |
45
+ | cover | Surface slips/coatings |
46
+ | ware | Manufacturing technique groups |
47
+ | form | Vessel shape/type |
48
+ |culcat|Cultural Category|
49
+
50
+ **Classes**: 9 pottery types
51
+
52
+ ## Training Methodology
53
+ - **Optimizer**: Adam
54
+ - **Learning Rate**: Initial 0.01 with exponential decay
55
+ - **Regularization**:
56
+ - He Weight Initialization
57
+ - L2 Regularization
58
+ - Early Stopping
59
+
60
+ ## Intended Use
61
+ - Archaeological ceramic type classification
62
+ - Research in archaeological artifact analysis
63
+ - Pottery provenance studies
64
+
65
+ ## Limitations
66
+ - Trained on a specific archaeological dataset
67
+ - Performance may vary with different ceramic collections
68
+ - Requires careful preprocessing of input features
69
+
70
+ ## Citation
71
+ If you use this model, please cite:
72
+ samanthajmichael/2025
73
+
74
+ ## License
75
+ MIT