Computer Vision
Pixel, classical filtering, feature detection အခြေခံမှ CNN architecture, transfer learning, object detection, segmentation, vision transformer, generative vision model အထိ သင်ပေးပြီး image classifier, fine-tuned classifier, object detector hands-on projects များပါဝင်သော Computer Vision tutorial ၂၅ ခန်း။
25 သင်ခန်းစာများ · 750 မိနစ် · Learning Platform Editorial Team
လက်ထဲမှာ ဘာရှိရင် ရပြီ
- PyTorch, torchvision install လုပ်ထားသော Python 3 (ဒီ course ရဲ့ example အားလုံးအတွက် CPU-only လုံလောက်သည်)
- Code editor (VS Code အကြံပြု)
- GPU (သို့) paid cloud account မလိုအပ်ပါ — example အားလုံး CPU ပေါ်မှာ run နိုင်သည်၊ ခဏနှေးမည်ပါ
မစခင် ဒီလောက်ပဲ သိထားပါ
- Python နှင့် PyTorch fundamentals — tensor, autograd, nn.Module, training loop, basic CNN တို့ကို ကြိုသိပြီးသားဟု ယူဆထားသည် (ဒီ site ရဲ့ Deep Learning with PyTorch course က ဒါတွေအားလုံးကို ဖုံးလွှမ်းထားသည်)
- ဒီ course က general deep learning mechanics ကို ပြန်မသင်ဘဲ vision-specific အကြောင်းအရာများ (image representation, classical image processing, vision-specific architecture/task) ကိုသာ နက်နက်ရှိုင်းရှိုင်း သင်ပေးမည်
- Computer vision (သို့) OpenCV အတွေ့အကြုံ မလိုအပ်ဘဲ စတင်နိုင်သည်
ဒီသင်တန်းပြီးရင် ကိုယ်ဘာလုပ်တတ်မလဲ
- Image တွေကို tensor အဖြစ် ဘယ်လို ကိုယ်စားပြုတယ်ဆိုတာ ရှင်းပြနိုင်ပြီး classical filtering, edge detection ကို အသုံးချနိုင်ခြင်း
- CNN-based image classifier များ ဆောက်/train/evaluate နိုင်ပြီး vision task အသစ်များအတွက် transfer learning အသုံးချနိုင်ခြင်း
- Object detection, semantic segmentation အခြေခံများ (IoU, mAP-style evaluation အပါအဝင်) ရှင်းပြနိုင်ခြင်း
- Modern vision architecture များ ရှင်းပြနိုင်ခြင်း — ResNet ရဲ့ skip connection, Vision Transformer, generative vision model
- Task တစ်ခုအတွက် မှန်ကန်တဲ့ architecture နှင့် evaluation metric ကို ရွေးချယ်နိုင်ခြင်း
- Image classifier, fine-tuned classifier, object detector လို real computer vision project ငယ်လေးများ ဆောက်/test နိုင်ခြင်း