一個帶有相機分類、伺服器分流器、料箱和校準流程的小型輸送機分類器，無需先進行完整 AI 訓練。
We may earn a commission from purchases through these links, at no extra cost to you.
物體分類是一個很好的機器人項目，因為環境是受限的：物體在輸送帶上到達，照明可以固定，料箱也是已知的。這比要求移動機器人在房間內識別所有物體要容易得多。
先從顏色或標記物分類開始，而不是直接訓練模型。一旦機械結構運行正常，再加入相機分類器。最大的可靠性提升點並非神經網絡；而是穩定的照明和物體間隔閘門，確保相機一次只看到一個物體。
Pulling current matching products on AliExpress…
Use HSV thresholds and colored blocks before training models.
Sort known clean objects like cans, caps and cardboard cards.
Count screws, washers or printed parts into bins using the same conveyor.
No. Start with color, shape or fiducial markers. Add ML after the conveyor is reliable.
Bad lighting, overlapping objects and motion blur cause most errors.
Slow enough for the camera exposure and servo gate. Reliability beats speed in the first version.
用預填 prompt 開始，等 RoboHub 生成即時零件清單、接線、CAD 同 firmware。