Cat detection using computer vision

Abstract

In this project, the author learned about object detection and utilised the object detection functions of the system to detect cats in digital images.

The author determine the effectiveness of the current deformable part model for detection of cat objects utilised by the system. This was done by evaluation of the detection results using a Java program that compares the ratio of missed detection, false positive detection as well as true positive detection to. It was determined that the current model was not effective enough as it was not able to reach the acceptable performance standard.

The author wrote a Java program to remedy the above-mentioned problem. The program handles rotated variants of an image with its detection results and converts them to display meaningful results on the un-rotated image.

The results of the program improved the detection performance of the deformable part model for detection of cat objects and reached the desired performance standard.

However the author believes that the algorithm could be enhanced to reduce processing time of images. The reliability of the results can also be tested with more and bigger sample sizes.

The author also recommended the expansion of the project scope. The detection results of rotated variants of an image could be synthesized to give a more accurate and comprehensive detection of the cat object in an image.

In addition, the evaluation process of the converted detection results by the Java program could also be made automated. This would allow for bigger sample sizes to be evaluated at a shorter time.