1 / 12

Android-Based Automatic Object Detection: A 2014 CV Project by Ohad Zadok

This project explores automatic object detection using an Android application to identify small objects in images. It aims to develop a reliable system that can be applied in security contexts, such as detecting enemy units in satellite images or identifying weapons in security camera feeds. The approach involves training an SVM model with image descriptors and validating it against new objects captured by the device's camera. While the model yields accurate results with small training sets, its reliability diminishes with larger datasets, highlighting the need for improved descriptors in future iterations.

tatiana-yeva
Télécharger la présentation

Android-Based Automatic Object Detection: A 2014 CV Project by Ohad Zadok

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Android Automatic object detection CV Project – 2014 By Ohad Zadok.

  2. Introduction Motivation Automatic Object Detection can be used to detect small objects in a photo and classify them. It can have security usages when trying to detect enemy units in a satellite images or weapons in security cameras. It can also be used to determine the context in which an image is taken by the objects in it.

  3. Introduction Project Goal The project goal is to successfully identify objects from a given set after training a model.

  4. Approach and Method Extracting Descriptors of an image to a vector of descriptors. Creating an SVM model from the set of images. Taking a photo of the new, unidentified object. creating again the same vector of descriptors and run the model on the vector. More on that at the Doc.

  5. Algorithms GrabCut http://grabcut.weebly.com/uploads/1/3/7/5/13757421/2018613.jpg?1351369992

  6. Algorithms GrabCut Before After

  7. Algorithms SVM http://www.thebookmyproject.com/cars/intrusion-detection-technique-using-k-means-fuzzy-neural-network-svm-classifiers/

  8. Implementation • Android System. • Photos from device's native camera. • A vector of descriptors was than calculated using opencv extension. • SVM model was created using LibSVM library. • The Entire project is run on a device with no need for external resources.

  9. Implementation

  10. Implementation

  11. Conclusion When applying the system to relatively small number of training sets the results are pretty accurate. When we apply it to a bigger set of objects it become unreliable, Future versions of the system can try to use different descriptors such as contours and get even more reliable results.

  12. References • http://www.csie.ntu.edu.tw/~cjlin/libsvm/ • https://en.wikipedia.org/wiki/Support_vector_machine • http://www.cs.ru.ac.za/research/g02m1682/

More Related