Team AGV just got a beast!

Finally, I would soon stop envying the research groups of the first world universities. My own research group AGV, headed by Prof Debashish Chakravarty is getting a huge overhaul. The best part, it all came as a surprise to all of us.

Until recently, we worked on our own laptops and the tasks related to IGVC did not require very high computational power as no high-level machine learning tasks were involved in navigating the robot. A Core i5 mobility laptop was more than sufficient for the same. But come Mahindra Rise problem statements, the situation is going to be different. A lot of high level data training is to be performed, including but not limited to traffic sign recognition, traffic light detection, bumper detection and pothole avoidance. All of them require neural network training in some form or the other, and hence a lot of time this semester (and the ones following) would involve training datasets for the same. And we now have the perfect machine for the same.

 

At its heart lies an Intel Xeon CPU, a whopping 32 gigabytes of RAM, a 250GB SSD for fast boots and 1TB HDD for storage. And the one thing we all were waiting for. The GPU we all adored is now with us, the Zotac GTX 970 with 4 GB video RAM! I installed the OS on the behemoth yesterday, and the boot times are astonishing to say the least. CUDA and ML libraries using the same would be installed soon, and the super fast GPU based Deep Learning training sessions would not be uncommon sights at late night AGV stints.

We all are too pumped up for unleashing the power of this behemoth on the mightiest of the datasets 😛 I’d keep posting good results on my blog. This is happening!!

 

Leave a Reply

Your email address will not be published. Required fields are marked *