the data from German Traffic Sign Dataset.
Download from this site, which had the image to 32×32.
data info:
- Number of training examples = 34799
- Number of validation examples = 4410
- Number of testing examples = 12630
- Image data shape = (32, 32, 3)
- Number of classes = 43
some data and labels:
only shuttle the data just like: X_train, y_train = shuffle(X_train, y_train)
the original network from LeNet-Lab Implement the LeNet-5 neural network architecture.
The LeNet architecture accepts a 32x32xC image as input, where C is the number of color channels. if are grayscale, C is 1 in this case. The channles in those image is 3(RGB)
Layer 1: Convolutional. The output shape should be 28x28x6. Activation. Your choice of activation function. Pooling. The output shape should be 14x14x6.
Layer 2: Convolutional. The output shape should be 10x10x16. Activation. Your choice of activation function. Pooling. The output shape should be 5x5x16.
Flatten. Flatten the output shape of the final pooling layer such that it's 1D instead of 3D. The easiest way to do is by using tf.contrib.layers.flatten, which is already imported for you.
Layer 3: Fully Connected. This should have 120 outputs. Activation. Your choice of activation function.
Layer 4: Fully Connected. This should have 84 outputs. Activation. Your choice of activation function.
Layer 5: Fully Connected (Logits). This should have 10 outputs.
Layer 1: Convolutional. The output shape should be 30x30x6. Activation. Your choice of activation function.
Layer 2: Convolutional. The output shape should be 28x28x6. Activation. Your choice of activation function. Pooling. The output shape should be 10x10x6.
Layer 3: Convolutional. The output shape should be 10x10x16. Activation. Your choice of activation function. Pooling. The output shape should be 5x5x16.
Flatten.
Layer 4: Fully Connected. This should have 120 outputs. Activation. Your choice of activation function. dropout. Layer 5: Fully Connected. This should have 84 outputs. Activation. Your choice of activation function.
Layer 6: Fully Connected (Logits). This should have 43 outputs. Final model just like this:
| Layer | Description |
|---|---|
| Input | 32x32x3 RGB image |
| Convolution 3x3 | outputs 30x30x6 |
| RELU | tf.nn.relu |
| Convolution 3x3 | outputs 28x28x6 |
| RELU | tf.nn.relu |
| Max pooling | 2x2 stride, outputs 14x14x6 |
| Convolution 5x5x16 | outputs 10x10x16. |
| RELU | tf.nn.relu |
| Max pooling | 2x2 stride, outputs 5x5x16 |
| Flatten | 1600. |
| RELU | tf.nn.relu |
| Flatten | 120. |
| RELU | tf.nn.relu |
| Flatten | 84. |
| RELU | tf.nn.relu |
| Dropout | 0.7 |
| Flatten | 43. |
Return the result of the 2nd fully connected layer.
Test Accuracy = 0.932 Validation Accuracy = 0.933 Train Accuracy = 0.992 more thing can read in report.html.
the origin data:
the lables: test_labels = [35,38,37,14,13]
the predict results: [35 38 37 1 13]
| Image | Prediction |
|---|---|
| Ahead only | Ahead only |
| Keep right | Keep right |
| Go straight or left | Go straight or left |
| Stop | Speed limit (20km/h) |
| Yield | Yield |
| the Output Top 5 Softmax Probabilities For Each Image: | |
 |