Classification and object localization of the handwritten digits

Table of Contents

• General Info
• Results
• Technologies Used
• Contact

General Information

• The project has two goals: 1) first to classifiy the handwritten digit 2) predict the bounding box around the hand-written digit
• I am using a branched neural network architecture using TensorFlow functional API:
  • 1. feature_extractor: these convolutional layers extract the features of the image.
  • 2. classifier: This define the output layer that predicts among 10 categories (digits 0 through 9)
  • 3. bounding_box_regression: This defines the output layer that predicts 4 numeric values, which define the coordinates of the bounding box (xmin, ymin, xmax, ymax)

• The link to the dataset: Handwritten digits dataset.

• The model achieves an accuracy of 99% for the classification and MSE of 0.0012 for the regression on the validation set after 10 epochs.
• The notebook has the capability to run on either TPUs, GPUs or CPUs. It uses tf.distribute.MirroredStrategy in case of multiple GPUs and tf.distribute.experimental.TPUStrategy in case of TPU.

Results

Technologies Used

• Python
• Tensorflow
• Pandas
• Matplotlib
• Keras

Contact

Created by Miralireza Nabavi - feel free to contact me!