Towards Network-accelerated ML-based Distributed Computer Vision Systems
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Computer vision is a crucial component in many modern applications (e.g., medical image analysis, environmental monitoring and self-driving cars). However, their stringent computational, latency and bandwidth requirements still pose a huge challenge to system architects, which must seek for alternatives to both the limited resources (e.g., low-end CPU) on client devices and the hurdles of moving data from clients to cloud/edge servers for analysis. In this work, we advocate for the usage of emerging programmable network devices to speed up ML-based computer vision tasks, particularly image classification, on resource constrained environments. To take the first step towards this new paradigm, we propose NetPixel, a framework that enables P4-programmable switches to classify images in realtime, accurately and at scale. We implemented a prototype of NetPixel in a software switch to show its feasibility and conducted a preliminary evaluation on widely adopted datasets. Our results show that NetPixel can classify images with an accuracy within 8% that of a server-based implementation even for shallow classifiers and low-resolution images.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it