EfficientDet: a new family of efficient object detectors. It is based on EfficientNet, and many times more efficient than state of art models. Link: https://arxiv.org/abs/1911.09070 Code: coming soonpic.twitter.com/2KYabAnpLL
Prikaži ovu nit
-
Can we deduct frame-per-second figures from Table 2 in the paper? As is unfortunately often the case, GFLOPs don't correlate to GPU performance at all. The 'latency' table seems more relevant but the numbers are strangely different from real world figures I'm familiar with.
Maybe GFLOPs correlate with energy consumption? Although memory use and transfers are probably also very energy intensive.
-
-
It's a more theoretical measure of efficiency, which correlates to performance when parallelism is less of an option. So for CPUs and edge devices it is usually actually very important. For GPUs - massively parallel but weak processors - it isn't a good predictor of performance.
-
You have a good point. Please take a look at my additional tweet above for latency on CPU and GPU (Figure 4 in the paper).
- Još 1 odgovor
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.