Comexp Launches TAPe+ML v3.5 Vision Software

The compact platform combines image classification, object detection and instance segmentation in one model that runs on a single CPU core.

Updated on Nov. 12, 2026 in Artificial Intelligence

Comexp Launches TAPe+ML v3.5 Vision Software

Comexp Research Lab has introduced TAPe+ML v3.5, a compact computer vision software platform that combines image classification, object detection and instance segmentation in a shared recognition core. The system has fewer than 100,000 parameters and a total file size of around 4.2MB.

Why it matters

The platform is designed to run without GPU acceleration, which could make computer vision practical on low-cost, low-power hardware. Its tested performance on inexpensive single-board and CPU systems suggests a path toward embedded and edge deployments.

The model has fewer than 100,000 parameters and runs at 8.6ms to 34.7ms per frame on one CPU core, with 20.4ms on Raspberry Pi 5 and 34.7ms on Raspberry Pi 4. Independent verification of these figures and head-to-head comparisons with rival models have not been reported.

The players

Comexp Research Lab

A research group that developed the TAPe+ML v3.5 computer vision platform and published its benchmark results.

The details

TAPe+ML builds a structured representation of each image before performing inference, and it uses a learned region selector on larger images. The platform has been tested on Raspberry Pi 4, Raspberry Pi 5 and Intel N100 CPUs, and the developer reports that it runs without GPU acceleration.

Timeline

  1. Introduced on November 12, 2026.

The Tech Race

The release fits a wider push to run vision models on inexpensive Raspberry Pi-class hardware rather than GPU servers, a shift that rival edge-AI developers are also pursuing. Benchmarks on those boards are increasingly the basis for comparing lightweight models.

Developers building cameras, robots or sensors on low-cost hardware could use a model of this size without buying a GPU. The developer's published benchmarks are the main evidence so far, so independent testing will determine how well the results hold up in real deployments.

The takeaway

Lightweight models that run on a single CPU core are narrowing the gap between capable computer vision and inexpensive hardware. Readers evaluating edge AI should check independent benchmarks before relying on a developer's reported figures.

Further reading

For more on the models powering systems like this, see Artificial Intelligence.

Source note: This article includes information reported by Electronics For You - Official Site ElectronicsForU.com.