wenmin-wu

wenmin-wu/ds-skills

Skill Claude CodeCodex

Encode text prompts into fixed-length dense vectors using SentenceTransformer for cosine-similarity evaluation in image-to-text retrieval tasks.

wenmin-wu/ds-skills

Skill Claude CodeCodex

2D CNN with asymmetric kernels — temporal convolutions (Nx1) then spectral convolutions (1xM) — to decouple time and feature extraction.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Track dice score separately for positive (mask-present) and negative (empty-mask) images to avoid division distortion.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Siamese network head that compares two embeddings via element-wise multiply, add, abs-diff, and squared-diff features for verification tasks.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Sigmoid-transformed normalized RMSE that maps error from [0,inf) to a bounded (0,1] similarity score using R2-score ratio.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Resample a 3D medical volume to a fixed depth N (e.g. 32) and feed the N slices as input channels to a 2D CNN with inchans=N instead of using a 3D conv backbone — gets the volumetric context for a fraction of the memory and lets you use any timm 2D pretrained model.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Encode the empirical per-position event rate as a soft prior over slice index by binning normalized slice location and looking up a precomputed weight vector — adds class-aware spatial context to a per-slice classifier without changing the model.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Pad 3D volumes with fewer slices than required by duplicating existing slices with slight brightness variation via convertScaleAbs.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Encode game state as a Super Mini Map (SMM) with separate binary channels for players, ball, and ownership, bit-packed for efficient transfer in RL training.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Cyclic cosine annealing LR that produces M diverse snapshots from a single training run for free ensembling.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Aggregate patch-level count-regression predictions into image-level totals by multiplying each prediction by a boolean mask (x (x > tau)) instead of rounding, retaining fractional evidence above threshold while killing background noise.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Turn sparse single-frame event labels (impacts, collisions, goals) into a usable detector training set by stamping the positive label onto a ±k-frame window around each event, then dropping any frame that contains no positives — gives the detector enough positive samples without changing the original annotation.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Split large-image segmentation data into train/val by spatial rectangle regions with border buffer exclusion to prevent patch leakage.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Compute normalized spectral band ratios (NDWI, CCCI, NDVI) from multispectral imagery and threshold for binary segmentation of water, vegetation, or other targets.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Keras callback implementing Stochastic Weight Averaging (SWA) — running mean of model weights over final training epochs.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Stratifies cross-validation folds by output sequence length to ensure balanced length distributions across train/val splits.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Normalize raw stroke coordinates to 0-255 range, resample at uniform arc-length spacing, then apply Ramer-Douglas-Peucker simplification.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Render stroke sequences to grayscale images with temporal intensity encoding where earlier strokes are brighter and later strokes fade to encode drawing order.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Serialize structured ground truth (chart data, table rows, form fields) as special-token-delimited sequences for generative vision-language model training.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Add random temporal offset to the center frame during training to augment temporal diversity in video-based models.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Pad batch sequence lengths to multiples of 8 for efficient tensor core utilization on GPUs, with -100 masking for label padding.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Compiles a PyTorch model to a TensorRT FP16 engine via torchtensorrt for 2-5x inference speedup, saved as reusable TorchScript.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Applies geometric and color augmentations at inference time and averages predictions to reduce variance.

wenmin-wu/ds-skills

Skill Claude CodeCodex

Assembles N extracted tiles into a sqrt(N) x sqrt(N) mosaic image for single-forward-pass CNN inference on whole slide images.