Research group on Embedded Architecture for Multisensing

  • Haddoc2 is a tool to automatically design FPGA-based hardware accelerators for convolutional neural networks (CNNs). Using a Caffe model, Haddoc2 generates a hardware description of the network (in VHDL-2008) which is constructor and device independent. Haddoc2 is built upon the principals of Dataflow stream-based processing of data, and, implements CNNs using a Direct Hardware Mapping approach, where all the actors involved in CNN processing are physically mapped on the FPGA.


  • CAPH is a domain-specific  language for describing and implementing stream-processing applications on reconfigurable hardware, such as FPGAs. CAPH generates VHDL code from high-level descriptions of signal or image processing applications. CAPH relies upon the actor/dataflow model of computation. Applications are described as networks of purely dataflow actors exchanging tokens through unidirectional channels and the behavior of each actor is defined as a set of transition rules using pattern matching.


  • GPStudio is a toolchain to solve low-level hardware concerns. The developer can then concentrate on the porting of algorithms into the target hardware architecture. By leveraging the modular architecture concept, available Intellectual Properties (IP) modules can be instantiated with standard interfaces between sensors, processing and communication blocks. In this way, a cross-platform IP library is proposed to improve code re-usability for a wide range of smart camera applications. Once the application is defined, GPStudio automatically generates the architecture and the glue code for the targeted board.


  • INAOE/DREAM Benchmark dataset is an odometry benchmark which consists of 130 monocular sequences provided with ground truth trajectories for all the sequences. The data was recorded at full frame rate (60 Hz) and, the ground-truth trajectory was obtained from a high-accuracy motion-capture system with eight high-speed tracking cameras (100 Hz).