You can download data from this link https://d17h27t6h515a5.cloudfront.net/topher/2017/February/5898cd6f_traffic-signs-data/traffic-signs-data.zip
Achievied accuracy around 96%
Train on 34799 samples
Epoch 1/30 34799/34799 [==============================] - 10s 286us/step - loss: 1.8930 - acc: 0.4808 - val_loss: 0.5692 - val_acc: 0.8365
Epoch 2/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.4983 - acc: 0.8466 - val_loss: 0.3246 - val_acc: 0.9104
Epoch 3/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.2855 - acc: 0.9135 - val_loss: 0.2394 - val_acc: 0.9348
Epoch 4/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.1926 - acc: 0.9406 - val_loss: 0.2302 - val_acc: 0.9394
Epoch 5/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.1488 - acc: 0.9550 - val_loss: 0.2184 - val_acc: 0.9423
Epoch 6/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.1168 - acc: 0.9647 - val_loss: 0.2276 - val_acc: 0.9386
Epoch 7/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.1011 - acc: 0.9686 - val_loss: 0.2097 - val_acc: 0.9440
Epoch 8/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.0800 - acc: 0.9739 - val_loss: 0.2097 - val_acc: 0.9428
Epoch 9/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.0681 - acc: 0.9793 - val_loss: 0.1818 - val_acc: 0.9519
Epoch 10/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0622 - acc: 0.9811 - val_loss: 0.1961 - val_acc: 0.9504
Epoch 11/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.0548 - acc: 0.9834 - val_loss: 0.1868 - val_acc: 0.9540
Epoch 12/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0476 - acc: 0.9849 - val_loss: 0.1835 - val_acc: 0.9559
Epoch 13/30 34799/34799 [==============================] - 7s 200us/step - loss: 0.0445 - acc: 0.9859 - val_loss: 0.2101 - val_acc: 0.9494
Epoch 14/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.0401 - acc: 0.9874 - val_loss: 0.1861 - val_acc: 0.9549
Epoch 15/30 34799/34799 [==============================] - 7s 200us/step - loss: 0.0348 - acc: 0.9888 - val_loss: 0.1880 - val_acc: 0.9560
Epoch 16/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.0352 - acc: 0.9884 - val_loss: 0.1856 - val_acc: 0.9571
Epoch 17/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.0323 - acc: 0.9896 - val_loss: 0.1857 - val_acc: 0.9572
Epoch 18/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0291 - acc: 0.9907 - val_loss: 0.1917 - val_acc: 0.9567
Epoch 19/30 34799/34799 [==============================] - 7s 198us/step - loss: 0.0274 - acc: 0.9909 - val_loss: 0.2023 - val_acc: 0.9566
Epoch 20/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0284 - acc: 0.9908 - val_loss: 0.1707 - val_acc: 0.9601
Epoch 21/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0241 - acc: 0.9917 - val_loss: 0.1939 - val_acc: 0.9584
Epoch 22/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0218 - acc: 0.9928 - val_loss: 0.2135 - val_acc: 0.9551
Epoch 23/30 34799/34799 [==============================] - 7s 201us/step - loss: 0.0210 - acc: 0.9933 - val_loss: 0.1906 - val_acc: 0.9604
Epoch 24/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0212 - acc: 0.9934 - val_loss: 0.1893 - val_acc: 0.9587
Epoch 25/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0175 - acc: 0.9943 - val_loss: 0.1878 - val_acc: 0.9605
Epoch 26/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0203 - acc: 0.9938 - val_loss: 0.2004 - val_acc: 0.9583
Epoch 27/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0180 - acc: 0.9939 - val_loss: 0.1876 - val_acc: 0.9607
Epoch 28/30 34799/34799 [==============================] - 7s 200us/step - loss: 0.0156 - acc: 0.9949 - val_loss: 0.1933 - val_acc: 0.9610
Epoch 29/30 34799/34799 [==============================] - 7s 200us/step - loss: 0.0158 - acc: 0.9951 - val_loss: 0.2096 - val_acc: 0.9578
Epoch 30/30 34799/34799 [==============================] - 7s 199us/step - loss: 0.0162 - acc: 0.9949 - val_loss: 0.1909 - val_acc: 0.9606
Test loss: 0.19094456916148936
Test accuracy: 0.9605700712494688