Image classification accuracy on the ImageNet dataset has been a barometer for progress in computer vision over the last decade. Several recent papers have questioned the degree to which the benchmark remains useful to the community [33, 3, 31, 42, 36], yet innovations continue to contribute gains to performance, with today's largest models achieving 90%+ top-1 accuracy. To help contextualize progress on ImageNet and provide a more meaningful evaluation for today's stateof-the-art models, we manually review and categorize every remaining mistake that a few top models make and provide insights into the long-tail of errors on one of the most benchmarked datasets in computer vision. We focus on the multi-label subset evaluation of ImageNet, where today's best models achieve upwards of 97% top-1 accuracy. Our analysis reveals that nearly half of the supposed mistakes are not mistakes at all, and we uncover new valid multi-labels, demonstrating that, without careful review, we are significantly underestimating the performance of these models. On the other hand, we also find that today's best models still make a significant number of mistakes (40%) that are obviously wrong to human reviewers. To calibrate future progress on ImageNet, we provide an updated multilabel evaluation set, and we curate ImageNet-Major1: a 68-example "major error" slice of the obvious mistakes made by today's top models--a slice where models should achieve near perfection, but today are far from doing so.

This work presents a novel approach to neural architecture search (NAS) that aims to increase carbon efficiency for the model design process. The proposed framework CE-NAS addresses the key challenge of high carbon cost associated with NAS by exploring the carbon emission variations of energy and energy differences of different NAS algorithms. At the high level, CE-NAS leverages a reinforcement-learning agent to dynamically adjust GPU resources based on carbon intensity, predicted by a time-series transformer, to balance energy-efficient sampling and energy-intensive evaluation tasks. Furthermore, CE-NAS leverages a recently proposed multi-objective optimizer to effectively reduce the NAS search space. We demonstrate the efficacy of CE-NAS in lowering carbon emissions while achieving SOTA results for both NAS datasets and open-domain NAS tasks. For example, on the HW-NasBench dataset, CE-NAS reduces carbon emissions by up to 7.22X while maintaining a search efficiency comparable to vanilla NAS. For open-domain NAS tasks, CE-NAS achieves SOTA results with 97.35% top-1 accuracy on CIFAR-10 with only 1.68M parameters and a carbon consumption of 38.53 lbs of CO2. On ImageNet, our searched model achieves 80.6% top-1 accuracy with a 0.78 ms TensorRT latency using FP16 on NVIDIA V100, consuming only 909.86 lbs of CO2, making it comparable to other one-shot-based NAS baselines.

Models are trained with Stochastic Gradient Descent with momentum equal to 0.9 [ We use a learning rate annealing scheme, decreasing the learning rate by a factor of 0.1 every 30 epochs. We train all models for 150 epochs. Then, we select the best learning rate and weight decay for each method and run 5 different seeds to report mean and standard deviation. We use the validation set of ImageNet to perform cross-validation and report performance on it. In section G we train the Augerino method on top of the Resnet-18 architecture.

We thank all the reviewers for their constructive comments. We will provide details in the final draft. MCUNet shows consistent improvement across different devices (F746, H743) and tasks (classification, detection). R1: Whether the overall network topology brings major improvement. R2: Why the auto-tuning in TVM fails to work on MCUs.

In this section, we expand on our observation that similarity between training samples is not binary. Consider the images shown in Figure 6. As a consequence, any similarity between the anchor image and the so-called'negative' examples is completely ignored. Further, all'positive' examples are considered to be The batch size is set to 16000. We train on 4 A100 GPUs.