Joseph

Well, got computer vision loaded into the RobotDog, unfortunately it's a bit sluggish, Qwen and OpenCV take about 5 seconds to process the image on my AI computer.

Hoping to take a stab at computer vision with my little robot dog project before my token reset. Pretty much letting AI build it all out. It's going to be using Qwen's 4 billion parameter model and OpenCV. Hoping it gets it finished, otherwise my reset is tomorrow!

I've got about $40 worth of electronics here. The robot dog was around $24, the ESP32 Display was $14 and the camera was $6. I'm probably going to hijack the dog's speaker but in the meantime I'm using a separate one. Currently I can take snapshots from the camera, use the display to show the camera and play sounds using my phone or my AI server. The idea is going to be to give this little robot dog sight and hearing and eventually have my AI computer "drive it". Hoping to find a distilled image recognition/computer vision system, but haven't done too much research on that yet. @karpathy has a Youtube course online as a worst case scenario but I'm not planning on building it myself and hoping there's already things out there in the wild. It's amazing what we're able to do modwise with things in this new age!

So I'm trying to get more into robotics and doing so by being as cheap as possible and using existing equipment. Yesterday I linked my ESP32 2.8" display with touch screen (~$14) to an ESP32 camera ($6) using mostly AI to create the firmware and linkage system. I added a speaker to the display board. Today's goal to start hashing out a server for my AI computer to be MITM. Short term goal, camera sends image to AI computer, it determines what's in frame and reports it back to display. I'm eventually going to attach this to a $20 robotic dog I bought that is pretty cool, I have logic analyzers to decipher the remote control and I'll probably do that next week. Eventual goal to make this robot dog have vision, a voice and hearing and my AI computer processing most of it.