Tmhacks22 ((install)) -

Taking advantage of the web's WebGPU capabilities, SignLens used TensorFlow.js to translate American Sign Language finger-spelling into text in real-time via a laptop webcam. What impressed judges wasn't just the 94% accuracy, but the latency—under 50ms. The team built a custom WebAssembly module to handle the matrix multiplication normally reserved for Python.

If you missed , you are likely waiting for the next iteration (tmhacks23 or tmhacks24). Based on the lessons of tmhacks22 , here is how you should prepare: tmhacks22

For the participants, the timeline was unforgiving. Taking advantage of the web's WebGPU capabilities, SignLens

Because if proved anything, it is that the best boundaries are the ones you build to break. If you missed , you are likely waiting

Two freshmen built a PWA (Progressive Web App) for rural bus riders. Since their town had no live bus tracking, they attached a $15 Raspberry Pi with an LTE shield to a family member's car, simulating a bus route. They then built a SMS gateway so you could text a number and get "The bus is 7 minutes away." It was scrappy, hardware-heavy, and deeply useful.