[01] // COURSE COMPUTER VISION

Neural Networks & Computer Vision.

Build a detector that works on your table and someone else’s. Same playbook as pool-ball ID, package orientation, and sports overlays.

8 modules · self-paced8 lessonsSelf-paced

About this course

Build a detector that works on your table and someone else’s. Same playbook as pool-ball ID, package orientation, and sports overlays.

End-to-end vision ML: object detection, synthetic Blender data, domain randomization, cloud training, and holdout testing in new environments.

What you will gain

  • Perception loop you can explain

    Map camera → model → action for robotics, sports, and QC. Know when rules beat neural nets.

  • A labeled dataset strategy

    Compare real capture vs synthetic Blender data. Auto-labels from the scene graph, not hand-drawn boxes for 10k frames.

  • Domain-randomized training

    Vary lighting, materials, camera ISP (exposure, blur, barrel distortion) so the model generalizes.

  • Cloud-scale training

    Burst render and train jobs on rented GPUs. Trade dollars for days on a single workstation.

  • Honest evaluation

    Hold out “brother’s room” environments. Failure gallery + metrics you can defend.

  • Capstone you can demo

    One domain end-to-end: dataset → trained model → inference in a script or thin app.

Skills you will gain

  • Perception loop you can explain
  • A labeled dataset strategy
  • Domain-randomized training
  • Cloud-scale training
  • Honest evaluation

Your instructor

Talodabi Faculty

Talodabi · Academy

Earn a certificate

Certificate of completion

Neural Networks & Computer Vision

Finish all lessons to unlock a Talodabi certificate you can add to LinkedIn or your portfolio.