Can some one recommend a good FPGA project based on following criterion? I have done some FPGA development for fun and looking for something more serious.
1. doable with virtex 2 pro FPGA (I dont want to spend more money buying newer boards at this time)
2. don't want to use HLS unless absolutely necessary
3. Fairly complex
4. Can be monetized - (can have some commercial value)
5. Better if the project is related to trending tech in 2018, eg: deep learning, vision, etc
1 and (4,5) seems incompatible to me. Furthermore Vivado is not compatible with Virtex 2 Pro. With the latest Xilinx Vivado tools you will be able to target 7-series and beyond (UltraScale, Ultrascale+).
Have a look to the minized (89$) which is based on the Zynq 7007S, single core ARM9 and 14400 6-inputLUTs and 28800 FF which is equivalent to a 2VP30 or 2VP40 (the LUTs on the 2VP are 4-input LUTs, the multiplier is a 18x18 compared to the 25x18 with pre-adder and post-adder of the Zynq, ...)
Now if you want to have a project in deep learning, a platform where you would have something to start on is the Ultra96 (http://ultra96.org/) which outperforms any Virtex 2 Pros in all directions!
My 2 cents
Video stream compressor.
At present, many old compression method patents are expired,
and these methods may be freely implemented in FPGA.
Well VII Pro is more than a decade old, but back then people have implemented sophisticated logic with it. The current trend is in AI. Deep learning, regular learning, Data analytics .. are the buzz words. To pick what to do is quite tricky. This AI is more of algorithms and particular processing for different situations is required. Also data size is another variable to consider. FPGA as pure logic has advantages and limitations. Even with the latest family of FPGAs the limitations of pure logic prevail. If the logic family has a CPU in it, the pure FPGA fabric could work as a co-processor releasing cpu of the repetitive work - easy to say on paper. The algorithm has to be broken up into smaller pieces and most time consuming but less intelligent tasks could be done in FPGAs and this could be monetised as an IP. For such small modules VII Pro system performance is still decent when used along with a cpu on chip. I wished to share specific information but could not.