-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathnotes_updated
47 lines (27 loc) · 3.58 KB
/
notes_updated
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
ffmpeg -i /home/gason/demo_ltts/cash_recognition/data/a.mp4 -vf fps=1 /home/gason/demo_ltts/cash_recognition/data/fifty/2_%d.jpg
python3 /home/gason/tensorflow/tensorflow/examples/image_retraining/retrain.py --bottleneck_dir=/home/gason/demo_ltts/cash_recognition/junk --how_many_training_steps 3000 --model_dir=/home/gason/demo_ltts/cash_recognition/models --output_graph=/home/gason/demo_ltts/cash_recognition/models/cash_recog.pb --output_labels=/home/gason/demo_ltts/cash_recognition/models/labels.txt --image_dir /home/gason/demo_ltts/cash_recognition/data/
python3 /home/gason/tensorflow/tensorflow/python/tools/optimize_for_inference.py --input=/home/gason/demo_ltts/cash_recognition/models/cash_recog.pb --output=/home/gason/demo_ltts/cash_recognition/models/optimized.pb --input_names="Mul" --output_names="final_result"
######################################
python3 /home/gason/tensorflow-for-poets-2/scripts/quantize_graph.py --input=/home/gason/demo_ltts/cash_recognition/models/optimized.pb --output=home/gason/demo_ltts/cash_recognition/models/round_optimized.pb --output_node_names=final_result --mode=weights_rounded
###################################################
sudo cp /home/gason/demo_ltts/cash_recognition/models/optimized.pb /home/gason/demo_ltts/cash_recognition/models/labels.txt /home/gason/demo_ltts/cash_recognition/tensorflow-classifier-android/app/src/main/assets/
streamlit run my_app.py
# test 1
sudo python /home/user/Desktop/varun/tensorflow/tensorflow/examples/image_retraining/retrain.py --bottleneck_dir=/home/user/Desktop/varun/cash_recognition/junk --model_dir=/home/user/Desktop/varun/cash_recognition/models --output_graph=/home/user/Desktop/varun/cash_recognition/models/cash_recog.pb --output_labels=/home/user/Desktop/varun/cash_recognition/models/labels.txt --image_dir /home/user/Desktop/varun/cash_recognition/data/ --print_misclassified_test_images --flip_left_right --random_crop=10 --random_brightness=10 --how_many_training_steps=3000 --learning_rate=0.001 --train_batch_size=64
2019-10-31 09:07:03.488300: Step 2999: Train accuracy = 98.4%
2019-10-31 09:07:03.488370: Step 2999: Cross entropy = 0.395660
2019-10-31 09:07:03.539998: Step 2999: Validation accuracy = 95.0% (N=100)
Final test accuracy = 87.9% (N=66)
=== MISCLASSIFIED TEST IMAGES ===
/home/user/Desktop/varun/cash_recognition/data/Ten/1_1.jpg hundred
/home/user/Desktop/varun/cash_recognition/data/FiveHundred/1_47.jpg ten1
/home/user/Desktop/varun/cash_recognition/data/Hundred1/1_11.jpg hundred
/home/user/Desktop/varun/cash_recognition/data/Hundred1/1_8.jpg fifty1
/home/user/Desktop/varun/cash_recognition/data/fifty/1_15.jpg ten
/home/user/Desktop/varun/cash_recognition/data/Twenty/1_46.jpg ten1
/home/user/Desktop/varun/cash_recognition/data/hundred/1_16.jpg ten
/home/user/Desktop/varun/cash_recognition/data/hundred/1_22.jpg fifty1
Converted 2 variables to const ops.
Very low accuracy, confused with 10/20
# test 2
sudo python /home/user/Desktop/varun/tensorflow/tensorflow/examples/image_retraining/retrain.py --bottleneck_dir=/home/user/Desktop/varun/cash_recognition/junk --model_dir=/home/user/Desktop/varun/cash_recognition/models --output_graph=/home/user/Desktop/varun/cash_recognition/models/cash_recog.pb --output_labels=/home/user/Desktop/varun/cash_recognition/models/labels.txt --image_dir /home/user/Desktop/varun/cash_recognition/data/ --print_misclassified_test_images --flip_left_right --random_crop=10 --random_brightness=10 --how_many_training_steps=4000 --learning_rate=0.001 --train_batch_size=128