Week 8: Labelling dataset

This week, we kept polishing the detail of our SSD model for object detection. For test purposes, we did not use large dataset but using the pictures that we downloaded from google image. We utilized the google_image_download API which can help us download the image with key words. We can download maximum 100 pictures at a time. After downloading the pictures, we deleted the unwanted pictures such as wrong file types or wrong objects in the image. This week, we only download a few hundreds of images of cellphones and desks/tables, because we need eventually identify where the object will be. For instance, "a cellphone is on the table." The picture labelling process is very time consuming. We need to label each pictures and save an annotation xml file associated with the original pictures.

Comments

Post a Comment

Popular posts from this blog

Week11: Meeting with Dr. Min and project revisit

Image
 For detecting the object, we tried to detect chairs in the pictures by running the model. Although there are only three chairs in the pictures, the model detects every frames and shows location points more than three. Therefore, we tried to exclude the overlapping frame. As a result, outcome of running with the modified model is only three location like below picture. Also, we met Dr. Min this week and discussed the progress of the course project. Although we are encountering many challenges, we made tough progress so far and moved the project forward. The most take away from the meeting with Dr. Min is that we were reminded that the key concept of this project is to understand what assistive technology is, what this technology is for, how we should consider when we want to design an assistive technology. Hence, for the final report, we may want to deliver a simple and clear design which can help the physically disabled people. In order to achieve this goal, we re-visit the p
Read more

Week12: Demo, User Study, Feedback, and Acknowledgement

Image
1. Demo We upload the video demonstrating our system at the youtube and below URL will direct you to the demo video. https://www.youtube.com/watch?v=49FvM7HNh3E 2.User study We conducted user study because we found the importance on investigating how the potential user find this system useful for decreasing the physical burden and how system itself is not burdening. For designing the user study, we put improtance on physical burden because that is our main things that should be addressed to solve the problem the people with physical impairment have. With the participants engagement in the user study, we could assess the needs in system from different perspective and validate the system. We also filmed and edited the process of user study, but we are going to keep it as private video to keep the privacy of participants. 3. Feedback We got the feedback after presentation. We tried to address the question expressed in some of the comments here.  Q. Is the
Read more

Week5: Preparing for proposal and presentation

We had a meeting today to finalize the proposal and presentation. For proposal, Nulee was responsible for the first chapter, Dongfang prepared for LR, and Eunsuh completed the method and approach. Also, we made the presentation slides today and we designated Eunsuh to present the project background, project goals, and LR. Dongfang will present method, plan, and member roles. I
Read more