The group has the necessary infrastructures to carry out its research and transfer tasks. These are fundamentally:
- Computer network made up of process servers, peripherals, and personal computers. In recent times the network of computing servers has been strengthened due to the intensification of our activities in the field of “Machine learning with deep neural networks”. The servers are powered by Intel Xeon processors up to a total of 368 cores (736 parallel processing threads), approximately 3.2 TB of RAM and 120 TB of total storage capacity. Additionally, 5 of the servers feature 18 high-performance graphics processors (GPUs), with a total of 61,696 Cuda cores of parallel processing and 192 GB of RAM. There are two RAID systems with SSD disk, with capacities of 19 TB and 6.5 TB, for fast access by the servers to the files on disk.
The servers with GPUs and the storage servers form a low latency 10Gbps ethernet network sharing a switch suitable for these speeds. This equipment is suitable for day-to-day use, but it is scarce when we train Deep Neural Network (DNN) algorithms or when we compete in international evaluations.
- Audio and video material and control, processing and visualization equipment for making multimodal recordings. Cameras with depth and infrared sensors. Specific cameras and lenses for artificial vision applications in industry.
- Acoustics laboratory composed of a semi-anechoic acoustic chamber and a “non-environment” type listening room with hardware and software equipment for the acquisition and multichannel reproduction of acoustic and vibratory signals, and equipment for the acquisition and reproduction of underwater acoustic signals.