Research

Counting MNIST

A simple synthetic dataset and baseline model for visual counting. The task is to count the number of even digits given a 100x100 image, each with up to 5 randomly chosen MNIST digits.

We use rejection sampling to ensure digits are separated by at least 28 pixels. Reproduced with details from Learning to count with deep object features.

Instructions

  • Clone repo: git clone [email protected]:fomorians/counting-mnist.git
  • Generate TFRecords: python -m counting_mnist.create_dataset
  • Train baseline: python -m counting_mnist.main

Results

Model Accuracy
Always Predict Zero Even Digits 33%
Uniform Count Predictions 12%
Baseline Model 85%

NOTE: This is not a competitive dataset. It’s intended to be simple place to start for validating ideas in counting models.

Reverse Image Search

Reverse Image Search

We created a lightweight reverse image search and wrote up the details in a three part series for Codementor. Writeup Read …


Counting MNIST

Counting MNIST

A simple synthetic dataset and baseline model for visual counting.


Receptive Field Calculator

Receptive Field Calculator

Tool to compute the effective receptive field of a convolutional layer.


Contact Us

Work with Us

We work with companies ranging from startups with interesting data to enterprises with large-scale challenges.

Featured Project

Counting MNIST

A simple synthetic dataset and baseline model for visual counting.