Authors: Maitreya Venkataswamy, Nathaniel Lanier, Adrian Lam

Review of the Problem and the Dataset

The goal of this project is to investigate methods to classify the frames of videos using neural networks, and specifically to investigate architectures that consider temporal information alongside the spatial information of the video frames. We aim to demonstrate that naively applying a standard convolutional neural network to each frame of a video separately is an inferior method to a model that can capture temporal patterns between the frames of the video.

The dataset that we use is the Rat Social Interaction dataset, which is a series of videos of two rats that…

Authors: Maitreya Venkataswamy, Adrian Lam, Nathaniel Lanier

Introduction to Convolutional-LSTM Layers

One application of deep learning and convolutional neural networks is in classifying the frames of a video. Time-series image data (which is a fancy way to say video data) has important spacial structure in the images that the convolutional layers use to increase the accuracy over a normal multi-layer perceptron. But another important aspect of this kind of data is the time structure of the frames. …

Authors: Maitreya Venkataswamy, Adrian Lam, Nathaniel Lanier


This is the final blog post documenting our progress in building a neural network to perform cassava leaf disease classification. We recall, from blog post 1, the importance of cassava plants as a food source in tropical regions, and the damage that viral diseases inflict on its yield. Because identification of these diseased plants traditionally requires manual inspection by trained agricultural experts, the diagnosis process is both expensive and inefficient. Our goal with this project is to build and train a neural network that can accurately classify the type of disease present in…

Authors: Maitreya Venkataswamy, Adrian Lam, Nathaniel Lanier

Building a Data Augmentation Pipeline

One of the biggest steps we’ve taken to boost the performance during training is to use data augmentation. Data augmentation is the process of making changes to the training data in real time during training in order to generate new “synthetic” data points. This effectively increases the size of the training data, which is important in deep learning since deep neural networks benefit from large amounts of training data. The original training data is stored in TFRecord files, which we load into a Tensorflow Dataset using the TFRecordDataset class. We then apply a…

Authors: Maitreya Venkataswamy, Adrian Lam, Nathaniel Lanier


In case the word “cassava” doesn’t ring any bells, it refers to a plant that is extensively cultivated in the tropics, and is a major food staple in these regions. However, despite its resilience to harsh conditions, cassava is often affected by viral diseases, which can inflict serious damage to crop yield. An important procedure in cassava cultivation is thus the identification of disease-ridden plants so that they can be promptly removed before the disease spreads further. This inspection process has traditionally been done manually by agricultural experts, but this is both costly…

Most robots operate in environments that are rife with disturbances. Things like wind, dynamic environments, even faults in the robot hardware, can cause it to deviate from a desired path. As a result, in order for a robot to operate effectively, it needs to be able to determine its location with high accuracy and precision. This process is called localization, and it enables a robot to operate confidently in an environment which is constantly undermining the robot’s ability to perform its task.

Modern localization techniques in robotics and other autonomous systems utilize sensors on the robot that allow it to…

Maitreya Venkataswamy

M.S. Student in Data Science @ Brown University, interested in Robotics, Aerospace Engineering, and anything simulated.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store