The problem Robocop solves

We aim to equip security systems and unmanned ground vehicles, which have grown in popularity as well as usefulness, with skills at par with humans so that people gain more confidence in using them for tasks at all levels- from courier delivery, to land patrolling to military use cases. As artificial intelligence is expanding its use cases, one of the most advantageous and immediate application is in security. This is our attempt to make surveillance as impregnable as possible with minimum human intervention.

All these skills can be integrated with the user’s existing security surveillance system like a CCTV camera as well.
We have concatenated three major skills:

1. Video captioning-The model pens down everything and anything it views through its camera. A very useful feature for both CCTV and unmanned vehicles. We provide real-time captioning as well as captioning an uploaded video. The captions are saved in a CSV file, which, being lighter on memory, can be saved for a longer time as compared to daily videos captured by the CCTV, which can’t be stored for more than three months.

2. Searching- Having the CSV file with accurate recordings of events, any keyword can be searched in it which will appear alongside of its date and time of occurring and the camera number which recorded the event. A very useful feature in analysing past events.

3. Depth Estimation-The third most important feature is depth estimation of objects in the robot’s field of view. With this, it can easily perform object detection as well as the distance of it from itself. More interestingly, for a moving object, the distance keeps updating and the Robo can infer the speed as well.

4. Line Tracking- Line tracking is a feature where the exact path followed by moving objects is determined by the model. When deploying Unmanned vehicles take for example drones for serious tasks like enemy movement in defense, or for a robot to move on a busy road not interfering with people.

Challenges we ran into

We ran into the following challenges:

1. We were not able to convert it into an installable software due to the time crunch, so we have just included a demo of the features which can be integrated with unmanned security vehicles and other security systems.
2. We plan to improve the accuracy of our video captioning model and include more emotions in it.
3. There is a lack of datasets suitable for our highly specific use cases