tezvyn:

Computer Vision

Image/video models, diffusion, OCR, multimodal

301 bites

Computer Vision87 sec read

Cross-attention in transformer VQA models

WHAT IT TESTS: vision-language fusion via attention. OUTLINE: text queries attend over image regions, learning alignment that grounds words to visual content. RED FLAG: confusing self-attention with cross-attention or saying it just concatenates features.

Computer Vision84 sec read

The data association problem in SLAM

WHAT IT TESTS: correspondence reliability in SLAM. OUTLINE: matching observations to landmarks, why wrong matches corrupt the map, robust techniques like RANSAC and descriptor matching. RED FLAG: treating it as trivial or ignoring outlier rejection.

Computer Vision86 sec read

Core components of visual SLAM

WHAT IT TESTS: SLAM system structure. OUTLINE: tracking estimates per-frame pose, mapping builds and refines the 3D map, loop closure detects revisits and corrects drift. RED FLAG: confusing tracking with mapping or omitting loop closure.

Computer Vision2 min read

Adapting a 2D CNN for video action recognition

WHAT IT TESTS: extending image models to temporal data. OUTLINE: run the 2D CNN per frame, pool features over time, optionally add two-stream or 3D conv. RED FLAG: assuming a single frame suffices or ignoring temporal information entirely.

Computer Vision86 sec read

Deploying segmentation on edge devices

WHAT IT TESTS: model compression for edge. OUTLINE: pick efficient architectures, apply INT8 quantization, distill from a large teacher. RED FLAG: only shrinking input resolution or naming one technique without the accuracy-versus-latency trade-off.

Computer Vision2 min read

Improving small object detection

WHAT IT TESTS: practical detector debugging. OUTLINE: raise input resolution and tile, use feature pyramids for high-res features, tune anchors and copy-paste augmentation. RED FLAG: only saying collect more data without addressing scale.

Computer Vision88 sec read

What is Bundle Adjustment and why is it tractable?

WHAT IT TESTS: structure-from-motion optimization. OUTLINE: jointly refine 3D points and camera poses by minimizing reprojection error, expensive due to many coupled parameters; sparsity of the Jacobian and the Schur complement make it tractable.

Computer Vision78 sec read

Why learn detection and description jointly like SuperPoint?

WHAT IT TESTS: learned versus classical keypoints. OUTLINE: a shared backbone jointly optimizes detection and description for matching, sharing computation and learning data-driven robustness instead of hand-crafted heuristics.

Computer Vision75 sec read

Feature detector vs feature descriptor.

WHAT IT TESTS: keypoint pipeline fundamentals. OUTLINE: a detector finds where interesting points are, a descriptor encodes the local appearance around each so points can be matched. RED FLAG: conflating the two or thinking one method only does detection.

Computer Vision84 sec read

How is an HDR radiance map constructed from exposures?

WHAT IT TESTS: HDR radiance recovery. OUTLINE: recover the inverse camera response function from corresponding pixels, linearize each exposure to radiance, then merge with confidence weights into a floating-point radiance map.

Computer Vision80 sec read

Design drift detection and retraining for a CV model.

WHAT IT TESTS: MLOps drift handling. OUTLINE: monitor input and prediction distributions plus delayed labels, detect data and concept drift, trigger an automated retrain-validate-deploy pipeline with versioning and rollback.

Computer Vision80 sec read

Design a cashierless retail checkout vision system.

WHAT IT TESTS: end-to-end CV system design. OUTLINE: multi-camera coverage, person tracking and re-ID, product detection plus action recognition for take or return, sensor fusion, edge-cloud split, and failure handling.

Computer Vision84 sec read

CPU vs GPU vs Edge TPU for inference.

WHAT IT TESTS: inference hardware trade-offs. OUTLINE: CPU is flexible but slow, GPU offers massive parallelism at high power, Edge TPU gives efficient low-power int8 inference but is constrained; choose by latency, power, cost, and model fit.

Computer Vision77 sec read

How do you train and evaluate on imbalanced defect data?

WHAT IT TESTS: handling class imbalance. OUTLINE: resampling, class weighting, focal loss, and anomaly framing for training; evaluate with precision, recall, PR-AUC, and F-beta, not accuracy.

Computer Vision72 sec read

How do you speed up a slow detection model?

WHAT IT TESTS: inference optimization techniques. OUTLINE: quantization, pruning, distillation, lighter backbones, and resolution or batching tweaks, each trading some accuracy or effort for speed.

Computer Vision78 sec read

Precision vs recall in object detection.

WHAT IT TESTS: precision and recall trade-offs. OUTLINE: precision is fraction of detections that are correct, recall is fraction of true objects found; prioritize recall for safety-critical detection, precision when false alarms are costly.

Computer Vision73 sec read

What data augmentations help small image datasets?

WHAT IT TESTS: regularization through augmentation. OUTLINE: apply label-preserving transforms like flips, crops, rotation, color jitter, and mixing to enlarge effective data and reduce overfitting.

Computer Vision79 sec read

How does focus stacking build an all-in-focus image?

WHAT IT TESTS: extended depth of field via focus stacking. OUTLINE: align frames, measure local sharpness as a focus metric, select or blend the sharpest source per region, then composite.

Computer Vision75 sec read

Why merge a burst instead of one long low-light exposure?

WHAT IT TESTS: burst denoising fundamentals. OUTLINE: short frames avoid motion blur and clipping while averaging cuts noise; align frames and merge robustly, rejecting misaligned regions to avoid ghosting.

Computer Vision81 sec read

How does a plenoptic camera enable post-capture refocus?

WHAT IT TESTS: light field imaging and refocusing. OUTLINE: light field records ray direction plus position via a microlens array, refocusing shears and integrates the 4D data to choose a virtual focal plane.