Skip to content

S
Sign Language Classifier
Commit f5e79840 authored by Abhishek Gupta's avatar Abhishek Gupta
Browse files
Options

tuned model

parent 192aae07
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 1692 additions and 1330 deletions
.ipynb_checkpoints/cnn1-checkpoint.ipynb
View file @ f5e79840
This diff is collapsed.
.ipynb_checkpoints/cnn2-checkpoint.ipynb
View file @ f5e79840
This source diff could not be displayed because it is too large. You can view the blob instead.
.ipynb_checkpoints/non-neural-checkpoint.ipynb
View file @ f5e79840
...@@ -308,7 +308,7 @@ ...@@ -308,7 +308,7 @@
} }
], ],
"source": [ "source": [
"utils.save_results(y_test, y_test_knn)" "utils.save_results(y_test, y_test_knn,\"knn\")"
] ]
}, },
{ {
......
cnn1.ipynb
View file @ f5e79840
This diff is collapsed.
cnn2.ipynb
View file @ f5e79840
This source diff could not be displayed because it is too large. You can view the blob instead.
models/alexNet.hd5
View file @ f5e79840
No preview for this file type
models/ann.hd5
View file @ f5e79840
No preview for this file type
models/cnn1_model.hd5
View file @ f5e79840
No preview for this file type
non-neural.ipynb
View file @ f5e79840
This source diff could not be displayed because it is too large. You can view the blob instead.
results/ANN_accuracy.png
View replaced file @ 192aae07 View file @ f5e79840
results/ANN_accuracy.png

33.9 KB | W: | H:

results/ANN_accuracy.png

34 KB | W: | H:

results/ANN_accuracy.png
results/ANN_accuracy.png
results/ANN_accuracy.png
results/ANN_accuracy.png
  • 2-up
  • Swipe
  • Onion skin
results/ANN_confusion.png
View replaced file @ 192aae07 View file @ f5e79840
results/ANN_confusion.png

61.2 KB | W: | H:

results/ANN_confusion.png

63 KB | W: | H:

results/ANN_confusion.png
results/ANN_confusion.png
results/ANN_confusion.png
results/ANN_confusion.png
  • 2-up
  • Swipe
  • Onion skin
results/ANN_loss.png
View replaced file @ 192aae07 View file @ f5e79840
results/ANN_loss.png

31.2 KB | W: | H:

results/ANN_loss.png

31.4 KB | W: | H:

results/ANN_loss.png
results/ANN_loss.png
results/ANN_loss.png
results/ANN_loss.png
  • 2-up
  • Swipe
  • Onion skin
results/ANN_scores.txt 0 → 100644
View file @ f5e79840
precision recall f1-score support
0 0.41 0.95 0.57 331
1 0.96 0.88 0.92 432
2 0.95 0.93 0.94 310
3 0.90 0.74 0.81 245
4 0.81 0.86 0.83 498
5 0.74 0.90 0.81 247
6 0.88 0.71 0.78 348
7 0.92 0.81 0.86 436
8 0.80 0.73 0.77 288
9 0.70 0.51 0.59 331
10 0.82 0.90 0.86 209
11 0.72 0.50 0.59 394
12 0.71 0.47 0.56 291
13 0.98 0.65 0.78 246
14 0.78 0.86 0.82 347
15 0.53 0.74 0.62 164
16 0.24 0.38 0.29 144
17 0.35 0.46 0.39 246
18 0.68 0.62 0.65 248
19 0.50 0.56 0.53 266
20 0.77 0.60 0.68 346
21 0.66 0.70 0.68 206
22 0.69 0.60 0.64 267
23 0.67 0.46 0.54 332
accuracy 0.70 7172
macro avg 0.71 0.69 0.69 7172
weighted avg 0.74 0.70 0.71 7172
results/alexnet_accuracy.png
View replaced file @ 192aae07 View file @ f5e79840
results/alexnet_accuracy.png

32 KB | W: | H:

results/alexnet_accuracy.png

46.6 KB | W: | H:

results/alexnet_accuracy.png
results/alexnet_accuracy.png
results/alexnet_accuracy.png
results/alexnet_accuracy.png
  • 2-up
  • Swipe
  • Onion skin
results/alexnet_confusion.png
View replaced file @ 192aae07 View file @ f5e79840
results/alexnet_confusion.png

52.7 KB | W: | H:

results/alexnet_confusion.png

53.3 KB | W: | H:

results/alexnet_confusion.png
results/alexnet_confusion.png
results/alexnet_confusion.png
results/alexnet_confusion.png
  • 2-up
  • Swipe
  • Onion skin
results/alexnet_loss.png
View replaced file @ 192aae07 View file @ f5e79840
results/alexnet_loss.png

30.9 KB | W: | H:

results/alexnet_loss.png

27.5 KB | W: | H:

results/alexnet_loss.png
results/alexnet_loss.png
results/alexnet_loss.png
results/alexnet_loss.png
  • 2-up
  • Swipe
  • Onion skin
results/alexnet_scores.txt 0 → 100644
View file @ f5e79840
precision recall f1-score support
0 0.53 0.95 0.68 331
1 0.95 0.93 0.94 432
2 0.99 0.89 0.94 310
3 0.96 0.59 0.73 245
4 0.88 1.00 0.94 498
5 1.00 0.91 0.96 247
6 1.00 0.69 0.82 348
7 0.94 0.89 0.91 436
8 0.52 1.00 0.69 288
9 1.00 0.80 0.89 331
10 1.00 0.93 0.96 209
11 0.97 0.74 0.84 394
12 0.79 0.82 0.80 291
13 1.00 0.83 0.91 246
14 0.93 0.85 0.89 347
15 1.00 0.99 0.99 164
16 0.77 0.86 0.81 144
17 1.00 0.48 0.65 246
18 0.71 0.68 0.70 248
19 0.70 0.82 0.75 266
20 1.00 0.53 0.70 346
21 0.85 1.00 0.92 206
22 0.78 0.96 0.86 267
23 0.69 0.82 0.75 332
accuracy 0.83 7172
macro avg 0.87 0.83 0.83 7172
weighted avg 0.88 0.83 0.84 7172
results/cnn1_model_accuracy.png
View replaced file @ 192aae07 View file @ f5e79840
results/cnn1_model_accuracy.png

28.5 KB | W: | H:

results/cnn1_model_accuracy.png

28.9 KB | W: | H:

results/cnn1_model_accuracy.png
results/cnn1_model_accuracy.png
results/cnn1_model_accuracy.png
results/cnn1_model_accuracy.png
  • 2-up
  • Swipe
  • Onion skin
results/cnn1_model_confusion.png
View replaced file @ 192aae07 View file @ f5e79840
results/cnn1_model_confusion.png

52.8 KB | W: | H:

results/cnn1_model_confusion.png

48 KB | W: | H:

results/cnn1_model_confusion.png
results/cnn1_model_confusion.png
results/cnn1_model_confusion.png
results/cnn1_model_confusion.png
  • 2-up
  • Swipe
  • Onion skin
results/cnn1_model_loss.png
View replaced file @ 192aae07 View file @ f5e79840
results/cnn1_model_loss.png

25.8 KB | W: | H:

results/cnn1_model_loss.png

25.7 KB | W: | H:

results/cnn1_model_loss.png
results/cnn1_model_loss.png
results/cnn1_model_loss.png
results/cnn1_model_loss.png
  • 2-up
  • Swipe
  • Onion skin
results/cnn1_model_scores.txt 0 → 100644
View file @ f5e79840
precision recall f1-score support
0 0.83 1.00 0.91 331
1 1.00 0.99 1.00 432
2 0.92 0.88 0.90 310
3 0.94 0.96 0.95 245
4 0.92 1.00 0.96 498
5 1.00 1.00 1.00 247
6 0.92 0.70 0.80 348
7 0.86 0.94 0.90 436
8 0.87 1.00 0.93 288
9 0.96 0.96 0.96 331
10 0.88 0.89 0.88 209
11 0.98 0.94 0.96 394
12 1.00 0.95 0.97 291
13 0.84 0.81 0.83 246
14 1.00 0.99 0.99 347
15 0.93 1.00 0.96 164
16 0.75 0.57 0.65 144
17 0.83 0.84 0.83 246
18 0.69 0.82 0.75 248
19 0.91 0.98 0.94 266
20 0.89 0.68 0.77 346
21 0.98 1.00 0.99 206
22 0.86 0.92 0.89 267
23 1.00 0.88 0.94 332
accuracy 0.91 7172
macro avg 0.91 0.90 0.90 7172
weighted avg 0.91 0.91 0.91 7172
results/cnn2_wo_aug_scores.txt 0 → 100644
View file @ f5e79840
precision recall f1-score support
0 0.98 1.00 0.99 331
1 1.00 1.00 1.00 432
2 1.00 1.00 1.00 310
3 1.00 1.00 1.00 245
4 0.98 1.00 0.99 498
5 1.00 1.00 1.00 247
6 0.94 0.95 0.94 348
7 0.96 0.95 0.96 436
8 0.94 1.00 0.97 288
9 1.00 0.98 0.99 331
10 0.91 1.00 0.95 209
11 1.00 1.00 1.00 394
12 1.00 0.93 0.96 291
13 1.00 1.00 1.00 246
14 1.00 1.00 1.00 347
15 1.00 1.00 1.00 164
16 0.98 0.72 0.83 144
17 0.98 0.96 0.97 246
18 0.92 0.83 0.88 248
19 0.93 0.99 0.96 266
20 0.95 1.00 0.97 346
21 1.00 1.00 1.00 206
22 0.93 1.00 0.96 267
23 1.00 0.94 0.97 332
accuracy 0.97 7172
macro avg 0.97 0.97 0.97 7172
weighted avg 0.98 0.97 0.97 7172
results/svm_scores.txt → results/cnn_aug_scores.txt
View file @ f5e79840
...@@ -6,7 +6,7 @@ ...@@ -6,7 +6,7 @@
3 1.00 1.00 1.00 245 3 1.00 1.00 1.00 245
4 1.00 1.00 1.00 498 4 1.00 1.00 1.00 498
5 1.00 1.00 1.00 247 5 1.00 1.00 1.00 247
6 1.00 0.99 1.00 348 6 1.00 1.00 1.00 348
7 1.00 1.00 1.00 436 7 1.00 1.00 1.00 436
8 1.00 1.00 1.00 288 8 1.00 1.00 1.00 288
9 1.00 1.00 1.00 331 9 1.00 1.00 1.00 331
...@@ -20,10 +20,10 @@ ...@@ -20,10 +20,10 @@
17 1.00 1.00 1.00 246 17 1.00 1.00 1.00 246
18 1.00 1.00 1.00 248 18 1.00 1.00 1.00 248
19 1.00 1.00 1.00 266 19 1.00 1.00 1.00 266
20 1.00 1.00 1.00 346 20 0.99 1.00 1.00 346
21 1.00 1.00 1.00 206 21 1.00 1.00 1.00 206
22 1.00 1.00 1.00 267 22 1.00 1.00 1.00 267
23 1.00 1.00 1.00 332 23 1.00 0.98 0.99 332
accuracy 1.00 7172 accuracy 1.00 7172
macro avg 1.00 1.00 1.00 7172 macro avg 1.00 1.00 1.00 7172
......
results/knn_scores.txt 0 → 100644
View file @ f5e79840
precision recall f1-score support
0 0.84 1.00 0.91 331
1 0.94 0.93 0.93 432
2 0.97 1.00 0.98 310
3 0.76 0.94 0.84 245
4 0.79 0.97 0.87 498
5 0.88 0.93 0.91 247
6 0.91 0.94 0.92 348
7 0.96 0.95 0.95 436
8 0.86 0.69 0.77 288
10 0.82 0.59 0.69 331
11 0.95 0.93 0.94 209
12 0.79 0.52 0.63 394
13 0.79 0.64 0.70 291
14 1.00 0.92 0.96 246
15 0.99 1.00 1.00 347
16 0.95 1.00 0.97 164
17 0.33 0.61 0.43 144
18 0.67 0.86 0.76 246
19 0.74 0.69 0.71 248
20 0.43 0.66 0.52 266
21 0.69 0.53 0.60 346
22 0.74 0.74 0.74 206
23 0.82 0.69 0.75 267
24 0.97 0.70 0.81 332
accuracy 0.81 7172
macro avg 0.82 0.81 0.80 7172
weighted avg 0.83 0.81 0.81 7172
results/svc_confusion.png
View replaced file @ 192aae07 View file @ f5e79840
results/svc_confusion.png

51.7 KB | W: | H:

results/svc_confusion.png

51.7 KB | W: | H:

results/svc_confusion.png
results/svc_confusion.png
results/svc_confusion.png
results/svc_confusion.png
  • 2-up
  • Swipe
  • Onion skin
results/svc_scores.txt 0 → 100644
View file @ f5e79840
precision recall f1-score support
0 0.93 1.00 0.96 331
1 1.00 0.99 1.00 432
2 0.88 0.99 0.93 310
3 0.95 1.00 0.97 245
4 0.94 0.99 0.97 498
5 0.76 0.83 0.80 247
6 0.94 0.90 0.92 348
7 0.97 0.95 0.96 436
8 0.82 0.90 0.85 288
10 0.81 0.66 0.73 331
11 0.87 1.00 0.93 209
12 0.84 0.73 0.78 394
13 0.90 0.67 0.77 291
14 0.95 0.85 0.89 246
15 1.00 1.00 1.00 347
16 1.00 0.99 1.00 164
17 0.33 0.61 0.43 144
18 0.72 0.81 0.76 246
19 0.84 0.69 0.76 248
20 0.62 0.64 0.63 266
21 0.79 0.62 0.69 346
22 0.64 0.80 0.71 206
23 0.84 0.81 0.83 267
24 0.85 0.76 0.80 332
accuracy 0.85 7172
macro avg 0.84 0.84 0.84 7172
weighted avg 0.86 0.85 0.85 7172
utils.py
View file @ f5e79840
...@@ -32,7 +32,7 @@ def save_model_history(model_history, model_name): ...@@ -32,7 +32,7 @@ def save_model_history(model_history, model_name):
def save_results(y_test, y_pred, model_name="model"): def save_results(y_test, y_pred, model_name="model"):
with open('./results/svm_scores.txt', 'w') as f: with open("./results/"+model_name+"_scores.txt", 'w') as f:
print(metrics.classification_report(y_test, y_pred), file = f) print(metrics.classification_report(y_test, y_pred), file = f)
print(metrics.classification_report(y_test, y_pred)) print(metrics.classification_report(y_test, y_pred))
# Confusion matrix # Confusion matrix
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment