Wirewalk Instituteof AI
mod-204 Intermediate Waitlist

Computer Vision Models

For engineers who need to choose, fine-tune, evaluate and deploy image models for classification, detection and segmentation in 2026.

The curriculum is being prepared and reviewed. Enrollment is not open yet.

About this course

Computer vision has shifted from training a network from scratch to adapting strong pretrained backbones. This course explains the two families underneath almost everything, convolutional networks and vision transformers, and the self-supervised and image-text pretraining (DINOv3, CLIP and SigLIP-style models) that makes small labeled datasets go a long way.

You then work task by task: transfer learning for classification, real-time detection with Ultralytics YOLO and DETR-style detectors such as RF-DETR, and segmentation including Meta's Segment Anything family, whose SAM 3 release added text-prompted concept segmentation. The course covers the unglamorous parts that decide success: annotation formats, augmentation, mAP and IoU done correctly, class imbalance, and error analysis. It ends with export to ONNX and TensorRT, edge latency, and video tracking. Model names and licenses are current as of October 2026; check each checkpoint's license before commercial use.

What you will be able to do

  • Explain how convolutional networks and vision transformers process images and where each is preferred
  • Fine-tune a pretrained backbone for classification using linear probing and full fine-tuning
  • Train and evaluate an object detector with correct annotation formats, augmentation and mAP measurement
  • Use promptable and text-prompted segmentation models to bootstrap labels and build segmentation pipelines
  • Apply zero-shot classification and retrieval with image-text embedding models
  • Diagnose model errors with confusion analysis, per-class metrics and dataset audits
  • Export a vision model to ONNX or TensorRT and meet a latency budget on server or edge hardware

Syllabus

6 modules · 18 lessons · about 15 hours

Module 1: How machines see

The two architectural families behind modern vision models.
3 lessons
  1. Convolutions, receptive fields and ResNet-style CNNs
  2. Vision transformers: patches, attention and when ViTs win
  3. Pretraining for vision: supervised, self-supervised DINO-style and image-text

Module 2: Image classification and embeddings

Adapting backbones to your classes and using embeddings directly.
3 lessons
  1. Transfer learning with timm and torchvision: linear probe versus fine-tune
  2. Zero-shot classification and image search with CLIP and SigLIP
  3. Augmentation, imbalance and small-data strategies

Module 3: Object detection

Finding and localizing objects in real time.
4 lessons
  1. Boxes, anchors, NMS and the shift to NMS-free detectors
  2. Training Ultralytics YOLO on a custom dataset
  3. DETR-family detectors and RF-DETR
  4. Measuring detectors: IoU, precision-recall and COCO mAP

Module 4: Segmentation

Pixel-level understanding, from semantic masks to promptable models.
3 lessons
  1. Semantic, instance and panoptic segmentation
  2. Segment Anything: point, box and text-concept prompts with SAM 3
  3. Using foundation models to accelerate annotation

Module 5: Data and evaluation discipline

The dataset work that decides whether a vision model succeeds.
2 lessons
  1. Annotation formats and label quality: COCO, YOLO and Pascal VOC
  2. Error analysis: per-class metrics, failure galleries and slice testing

Module 6: Deploying vision models

Making models fast enough for production and handling video.
3 lessons
  1. Export to ONNX and TensorRT, FP16 and INT8
  2. Edge targets and latency budgets
  3. Video pipelines: tracking with ByteTrack and frame sampling

Exam and certificate

The exam unlocks once you have completed every lesson. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 45 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.

Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.

The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by PyTorch (torchvision), Ultralytics YOLO, Meta (DINOv3, SAM 3), Roboflow (RF-DETR, Supervision), Hugging Face (Transformers, timm), OpenAI (CLIP), Google (SigLIP), NVIDIA TensorRT, or ONNX Runtime. Product and company names are trademarks of their respective owners.

More in Model Development

All 16 courses in this track
mod-201 Intermediate Available

Deep Learning with PyTorch

A hands-on course in modern PyTorch: tensors and autograd, nn.Module design, data loading, a training loop you understand line by line, GPU and mixed-precision training, torch.compile, and exporting a model for deployment.

14 h · 19 lessons See price at checkout
mod-202 Intermediate Available

Transformers from the Ground Up

Build a decoder-only transformer piece by piece: tokenization, embeddings, scaled dot-product attention, multi-head and grouped-query attention, positional schemes such as RoPE, and the training and sampling loop, then map it onto real open-weight model architectures.

16 h · 19 lessons See price at checkout
mod-203 Intermediate Waitlist

Classical Machine Learning in Production

Gradient-boosted trees and linear models still win most tabular problems. This course covers leakage-proof pipelines, honest validation, tuning, calibration and explainability, then packaging, serving and monitoring for drift.

14 h · 19 lessons Join the waitlist