Cognitive Machines

In this demo we showcase the low latency vehicle counting use case for traffic analysis and has the following key features

* Vehicle make and model recognition

* Vehicle type classification

* Traffic pattern analysis

The customer introduced themselves with an interesting problem of analysing CCTV video recordings of traffic videos on roadways. They wanted to leverage the CCTV video recordings and generate insightful analytics about the traffic on residential and commercial properties. The challenge was to generate the results of the video analytics in a time-efficient and a power-efficient manner.

We developed a Deep Neural Network model to recognise six categories of vehicles namely cars, motorcycles, buses, trucks, bicycles and pedestrians. We first collected the CCTV footage and generated a customised data set containing 2000 images for each class of vehicle. We then trained a Deep Neural Network to classify each class of vehicle and localize them on a video frame. Since this was a video processing application, performance was a key metric in addition to the accuracy of the vehicle finder.

Manufacturing industries continuously monitor their products and artifacts for their precision and quality. It is imperative that they detect defects during their manufacturing process as this eliminates wastage in terms of raw material and the time to produce the same. It also paves way for good quality control during the production process. The client had a heavy industry manufacturing plant and they wanted to improve their manufacturing process by adding a customised machine vision system. They wanted us to build a system that can measure the dimensions and detect defects on the manufactured artifacts. In order to achieve this, we were required to study the manufacturing process and product based on their metrics of manufacturing, namely length, thickness, surface texture etc.

The key goals of the project are to identify the object of interest on the camera feed and to measure the object. Since the object of interest is not a standard shape, conventional shape recognition methods like Hough Transforms are not suitable for object detection. So we started the data acquisition phase of the project by collecting the video recording from the camera feeds and trim the video segments pertaining to the object of interest. This was then followed by an object annotation phase with the CVAT tool to generate a segmentation dataset of 5000 images. We implemented a neural network training process derived from the CNN architectures including the Xception, Inception and ResNet. From our RoC studies, it was observed that the ResNet model was performing adequately in terms of both accuracy and execution time. We then ported the ResNet inference program to an edge computing platform that interfaces along with the existing camera module. The edge computing platform read the camera feed and performed real time object measurement and reported the measurements over a network connection

Accounting is an integral part of business finance management. Keeping track of manual bills is an exhausting task and errors are likely to be introduced while handling large numbers of the same. Our client is an accounting company who wanted to automate some parts of their bills reconciliation process. Automation will help them perform the accounting quicker with less or no errors. The customers of our client will submit the manual bills for the accounting. The client wanted to build an OCR system to convert expense receipt stubs stored as scanned documents and images. In order to achieve this, we were requirement to extract elements and fields from the expense receipt stub, namely, date, total price, tax etc

In this demo, we identity individual pedestrians as seen on a video footage and track the movements of the pedestrian

* Pedestrian tracking

* Pedestrian counting

* Safe distance validation

This Android application showcases the features of the Computer Vision library developed at Cognitive Machines. The features include

• Pre processing filters
• Shape finders
• Feature detectors

The customer approached us in search of an efficient way to manage their raw timber warehouse. The existing system involved tedious manual counting of the multi-sized logs which exacerbated the painful process of maintaining daily records. The challenge here was to design and develop an effective system to automate the counting with accuracy and correctness.

We came up with a Computer Vision solution that can automatically count timber logs in a single image and generate a report on the inventory. First, we developed a Machine Learning system for training of timber logs based on common properties like shape, colour, size etc. The trained model was then deployed on a mobile application that allowed the warehouse inspector to take a picture of the timber pile and generate a report on the number of logs, size of the cluster etc