frereit / tensorflowhandwritingrecognition

Using Tensorflow to classify the NIST Dataset 19 (Handwriting)

Frederik (he/him)

• 👨‍🎓 Studying Cyber Security @ DHBW Mannheim (WS2021)
• 🏢 Employed @ cirosec
• 👨‍💻 Hacking @ DU4L
• 📫 [email protected] (:lock: S/MIME Certificate)

tensorflowhandwritingrecognition's Issues

trained model for capital letters

can you please share the trained model for capital letters. It will be very helpful to me.

The model used in this experiment

Hello,

Can you please the model used here so we can test and see the predictions without having to train the model from the start.

Thanks in advance

Question

Hi just want to check how are you matching the labels to the training data?

Like for e.g. inside class 4e, it contains letter 'N', so why is the label 4e instead of N?

How

The

accuracy only about 4% after 10k epochs

accuracy does not increase output of python training_32x32.py 20000
2018-04-14T05:55:41.833947 Variables initialized! 2018-04-14T05:55:41.833979 Training for 20000 epochs. 2018-04-14T05:56:02.271397 Step: 0; Accuracy: 0.065; 2018-04-14T05:56:40.071369 Step: 200; Accuracy: 0.039999995; Time (200 Steps): 37.7999420166 2018-04-14T05:57:17.666085 Step: 400; Accuracy: 0.065; Time (200 Steps): 37.594686985 2018-04-14T05:57:55.162879 Step: 600; Accuracy: 0.024999999; Time (200 Steps): 37.496762991 2018-04-14T05:58:32.752867 Step: 800; Accuracy: 0.06; Time (200 Steps): 37.5899569988 2018-04-14T05:59:10.389535 Step: 1000; Accuracy: 0.054999996; Time (200 Steps): 37.6366379261 2018-04-14T05:59:47.704203 Step: 1200; Accuracy: 0.044999994; Time (200 Steps): 37.3146338463 2018-04-14T06:00:24.949550 Step: 1400; Accuracy: 0.059999995; Time (200 Steps): 37.2453179359 2018-04-14T06:01:02.509568 Step: 1600; Accuracy: 0.049999997; Time (200 Steps): 37.5599889755 2018-04-14T06:01:39.896241 Step: 1800; Accuracy: 0.065; Time (200 Steps): 37.3866410255 2018-04-14T06:02:17.152708 Step: 2000; Accuracy: 0.054999992; Time (200 Steps): 37.2564370632 2018-04-14T06:02:56.307966 Step: 2200; Accuracy: 0.06; Time (200 Steps): 39.1552290916 2018-04-14T06:03:33.660739 Step: 2400; Accuracy: 0.04; Time (200 Steps): 37.3527429104 2018-04-14T06:04:10.871626 Step: 2600; Accuracy: 0.06999999; Time (200 Steps): 37.2108578682 2018-04-14T06:04:48.515811 Step: 2800; Accuracy: 0.049999997; Time (200 Steps): 37.6441559792 2018-04-14T06:05:25.962051 Step: 3000; Accuracy: 0.065; Time (200 Steps): 37.4462118149 2018-04-14T06:06:03.417133 Step: 3200; Accuracy: 0.07; Time (200 Steps): 37.4550538063 2018-04-14T06:06:40.669024 Step: 3400; Accuracy: 0.06; Time (200 Steps): 37.2518620491 2018-04-14T06:07:17.979902 Step: 3600; Accuracy: 0.029999997; Time (200 Steps): 37.3108501434 2018-04-14T06:07:55.542385 Step: 3800; Accuracy: 0.06; Time (200 Steps): 37.5624539852 2018-04-14T06:08:32.736097 Step: 4000; Accuracy: 0.059999995; Time (200 Steps): 37.1936819553 2018-04-14T06:09:11.440181 Step: 4200; Accuracy: 0.044999998; Time (200 Steps): 38.7040550709 2018-04-14T06:09:50.294338 Step: 4400; Accuracy: 0.034999996; Time (200 Steps): 38.8540999889 2018-04-14T06:10:28.694038 Step: 4600; Accuracy: 0.015; Time (200 Steps): 38.3996651173 2018-04-14T06:11:06.902555 Step: 4800; Accuracy: 0.08; Time (200 Steps): 38.2084801197 2018-04-14T06:11:45.240362 Step: 5000; Accuracy: 0.054999996; Time (200 Steps): 38.3377759457 2018-04-14T06:12:22.930590 Step: 5200; Accuracy: 0.035; Time (200 Steps): 37.6901938915 2018-04-14T06:13:00.311706 Step: 5400; Accuracy: 0.055; Time (200 Steps): 37.3810811043 2018-04-14T06:13:38.065907 Step: 5600; Accuracy: 0.029999997; Time (200 Steps): 37.7541618347 2018-04-14T06:14:15.843873 Step: 5800; Accuracy: 0.02; Time (200 Steps): 37.7779290676 2018-04-14T06:14:53.553676 Step: 6000; Accuracy: 0.044999998; Time (200 Steps): 37.7097699642 2018-04-14T06:15:31.632643 Step: 6200; Accuracy: 0.045; Time (200 Steps): 38.0789370537 2018-04-14T06:16:09.375511 Step: 6400; Accuracy: 0.065; Time (200 Steps): 37.7428379059 2018-04-14T06:16:46.825972 Step: 6600; Accuracy: 0.02; Time (200 Steps): 37.4504299164 2018-04-14T06:17:24.177718 Step: 6800; Accuracy: 0.03; Time (200 Steps): 37.3517150879 2018-04-14T06:18:02.276903 Step: 7000; Accuracy: 0.074999996; Time (200 Steps): 38.0991539955 2018-04-14T06:18:40.414380 Step: 7200; Accuracy: 0.065; Time (200 Steps): 38.1374459267 2018-04-14T06:19:18.211176 Step: 7400; Accuracy: 0.059999995; Time (200 Steps): 37.7967560291 2018-04-14T06:19:55.825618 Step: 7600; Accuracy: 0.044999994; Time (200 Steps): 37.6143639088 2018-04-14T06:20:33.420122 Step: 7800; Accuracy: 0.06; Time (200 Steps): 37.5944728851 2018-04-14T06:21:10.900863 Step: 8000; Accuracy: 0.054999996; Time (200 Steps): 37.4807109833 2018-04-14T06:21:49.031777 Step: 8200; Accuracy: 0.075; Time (200 Steps): 38.1308829784 2018-04-14T06:22:27.200214 Step: 8400; Accuracy: 0.065; Time (200 Steps): 38.1684019566 2018-04-14T06:23:05.441150 Step: 8600; Accuracy: 0.07; Time (200 Steps): 38.2408931255 2018-04-14T06:23:43.310775 Step: 8800; Accuracy: 0.07; Time (200 Steps): 37.869592905 2018-04-14T06:24:21.684700 Step: 9000; Accuracy: 0.01; Time (200 Steps): 38.3738861084 2018-04-14T06:24:59.754901 Step: 9200; Accuracy: 0.035; Time (200 Steps): 38.0701658726 2018-04-14T06:25:37.858305 Step: 9400; Accuracy: 0.06; Time (200 Steps): 38.103372097 2018-04-14T06:26:15.642299 Step: 9600; Accuracy: 0.055; Time (200 Steps): 37.783962965 2018-04-14T06:26:53.406536 Step: 9800; Accuracy: 0.059999995; Time (200 Steps): 37.7642059326 2018-04-14T06:27:31.237817 Step: 10000; Accuracy: 0.045; Time (200 Steps): 37.8312489986 2018-04-14T06:28:08.976475 Step: 10200; Accuracy: 0.044999998; Time (200 Steps): 37.7386269569 2018-04-14T06:28:46.723300 Step: 10400; Accuracy: 0.059999995; Time (200 Steps): 37.746792078 2018-04-14T06:29:24.354060 Step: 10600; Accuracy: 0.04; Time (200 Steps): 37.6307280064 2018-04-14T06:30:01.777509 Step: 10800; Accuracy: 0.089999996; Time (200 Steps): 37.423418045 2018-04-14T06:30:39.582372 Step: 11000; Accuracy: 0.039999995; Time (200 Steps): 37.804831028 2018-04-14T06:31:17.012337 Step: 11200; Accuracy: 0.074999996; Time (200 Steps): 37.4299340248 2018-04-14T06:31:54.561577 Step: 11400; Accuracy: 0.03; Time (200 Steps): 37.5492069721 2018-04-14T06:32:31.895916 Step: 11600; Accuracy: 0.024999999; Time (200 Steps): 37.334307909 2018-04-14T06:33:09.877085 Step: 11800; Accuracy: 0.075; Time (200 Steps): 37.9811351299 2018-04-14T06:33:47.663098 Step: 12000; Accuracy: 0.065; Time (200 Steps): 37.7859518528 2018-04-14T06:34:25.230280 Step: 12200; Accuracy: 0.085; Time (200 Steps): 37.5671510696 2018-04-14T06:35:02.982136 Step: 12400; Accuracy: 0.055; Time (200 Steps): 37.7518231869 2018-04-14T06:35:40.522167 Step: 12600; Accuracy: 0.044999994; Time (200 Steps): 37.5399930477 2018-04-14T06:36:18.259583 Step: 12800; Accuracy: 0.074999996; Time (200 Steps): 37.7373740673 2018-04-14T06:36:55.826241 Step: 13000; Accuracy: 0.049999997; Time (200 Steps): 37.5666291714 2018-04-14T06:37:33.069297 Step: 13200; Accuracy: 0.059999995; Time (200 Steps): 37.2430272102 2018-04-14T06:38:10.404923 Step: 13400; Accuracy: 0.049999997; Time (200 Steps): 37.3355939388 2018-04-14T06:38:47.898836 Step: 13600; Accuracy: 0.06; Time (200 Steps): 37.4938759804 2018-04-14T06:39:25.220859 Step: 13800; Accuracy: 0.035; Time (200 Steps): 37.3219900131 2018-04-14T06:40:02.689004 Step: 14000; Accuracy: 0.065; Time (200 Steps): 37.4681141376 2018-04-14T06:40:40.509674 Step: 14200; Accuracy: 0.06; Time (200 Steps): 37.8206310272 2018-04-14T06:41:17.805861 Step: 14400; Accuracy: 0.049999997; Time (200 Steps): 37.2961490154 2018-04-14T06:41:55.553839 Step: 14600; Accuracy: 0.044999998; Time (200 Steps): 37.7479400635 2018-04-14T06:42:33.157692 Step: 14800; Accuracy: 0.049999997; Time (200 Steps): 37.6038210392 2018-04-14T06:43:11.098131 Step: 15000; Accuracy: 0.044999998; Time (200 Steps): 37.9403998852 2018-04-14T06:43:48.738506 Step: 15200; Accuracy: 0.04; Time (200 Steps): 37.6403419971 2018-04-14T06:44:26.167303 Step: 15400; Accuracy: 0.065; Time (200 Steps): 37.4287588596 2018-04-14T06:45:03.604250 Step: 15600; Accuracy: 0.02; Time (200 Steps): 37.4369161129 2018-04-14T06:45:41.133063 Step: 15800; Accuracy: 0.055; Time (200 Steps): 37.5287799835 2018-04-14T06:46:18.581898 Step: 16000; Accuracy: 0.044999994; Time (200 Steps): 37.4488039017 2018-04-14T06:46:56.316838 Step: 16200; Accuracy: 0.07; Time (200 Steps): 37.7349050045 2018-04-14T06:47:33.987881 Step: 16400; Accuracy: 0.054999996; Time (200 Steps): 37.6710109711 2018-04-14T06:48:11.827668 Step: 16600; Accuracy: 0.034999996; Time (200 Steps): 37.8397481441 2018-04-14T06:48:49.399726 Step: 16800; Accuracy: 0.075; Time (200 Steps): 37.5720191002 2018-04-14T06:49:26.892750 Step: 17000; Accuracy: 0.059999995; Time (200 Steps): 37.4929881096

Datasets?

Hi,

It is possible for you to share labels and image files(.npy) and trained classifier on NIST files. It will be a great help for me. Currently, I figured out how to get the character out of the word, but CPU is not helping me train on the entire dataset.

Thanks in advance.

Sagar

Issue in Training 32x32 file!!

@frereit I got an error :

File "F:\TensorflowHandwritingRecognition-master\training_32x32.py", line 211, in
main()
File "F:\TensorflowHandwritingRecognition-master\training_32x32.py", line 184, in main
epochs = int(sys.argv[1])
IndexError: list index out of range

Can you please help me with this asap.

Predicting a full word

Hi, how can we we predict a full word using the same. Given an example of word image.

How to use trained model to predict new image with characters?

First thanks for your project. I was wondering since the models are trained on all english characters how I can predict using trained model on new english character image

which will give me out put text as sagar

Thanks