JobCam is a fun new way to find jobs in your area. It allows you to target your phone at an object, and then through the InfoJobs API we can find a job related to that object. The app will process data in real time allowing the user to move the phone around the room to find interesting jobs that they had not considered.
One of the main goals of our hack was to do real time object detection on a mobile phone. In this case were able to get fairly close to that goal. We offload the processing of each frame to the Object Detection Server (ODS) that is running Yolo (You Only Look Once), a very large convolutional neural network that can detect up to 80 different objects. Each frame is sent over a http request and the response returns a bounding boxes and classification of each object in the scene. Running a fairly powerful GPU on the ODS helps get close to real time classification on Android.
Each frame takes around 500-1000ms for a round trip, but we believe we could get much quicker response time using a lower level camera API than the one we are using. This is currently where the main bottleneck is and could be resolved given a day or two of more hacking.
With this cool tech in hand we attempted the InfoJob challenge, to see if we could bring real time object detection and their API together. We looked at doing an app where it allows you to target your phone at an object, and then through the InfoJobs API we can find a job related to that object. The app will process data in real time allowing the user to move the phone around the room to find interesting jobs that they had not considered before.
Although not the most practical job finding tool, we find this to be a fun activity and more engaging with customers with signing up and registering for a job. It is a novel way to search through their current job listings.
Log in or sign up for Devpost to join the conversation.