#efficientnet
An Improved EfficientNet-B0 Model for Facial Emotion Recognition #NeuroDegeneration 🧪🧠
https://www.researchsquare.com/article/rs-11679881/latest
November 11, 2026 at 8:02 AM
Shkelqim Sherifi: Within-Family EfficientNet Scaling under Fixed-Resolution Transfer Learning for Spectral-Domain Optical Coherence Tomography Retinal Disease Classification https://arxiv.org/abs/2611.06452 https://arxiv.org/pdf/2611.06452 https://arxiv.org/html/2611.06452
November 5, 2026 at 6:41 AM
さらに、画像識別 AI(Multi-branch EfficientNet)を用いてシルエットのみからグループを再現させたところ、約 89% の高い精度を記録し、判断根拠の可視化(Grad-CAM)によって AI が歯の頂部や縁のカーブを正しく認識していることが確認された。
本手法は、大量の化石データを客観的かつ高速にスクリーニングする次世代の恐竜研究ツールとして期待されれる。
一方、この研究の限界として
1)入力画像が物理的スケールに調整(scale-calibrated)されていない点、
(2)サンプルサイズ(64 標本)が機械学習としては比較的小規模である点、
October 28, 2026 at 11:13 PM
Future-dated timestamps, no variant, no evals: this Modelers.cn entry is a licensed placeholder, not a release. The real draw is Apache-2.0 weights — one git clone inside the Ascend ecosystem. Hard to see this beating upstream in production until someone explains '-es'.
EfficientNet Lands on Modelers.cn: A Compact Vision Backbone With a Clean Licence and an Empty Model Card
If you haven't come across Modelers.cn before, it's Huawei's model hub for the Ascend AI ecosystem — a place where Chinese developers publish and pull models that often don't surface on the usual…
aichina.news
October 19, 2026 at 11:12 AM
A new arXiv paper introduces HyMLRaman, a hybrid framework combining EfficientNet-B3 embeddings with generative feature expansion to boost Raman spectroscopy identification of six common drug compounds.

#AI #RamanSpectroscopy #MachineLearning #Pharma
https://arxiv.org/abs/2610.02224
October 5, 2026 at 10:02 AM
5) DINOv2 is a weak natural-image baseline for morphology, convnets (e.g. EfficientNet) should capture multi-scale texture better
September 11, 2026 at 7:25 AM
Sejuti Basu, Ashima Sood, Vijay Kumar, Sahil Sharma: Swin Meets EfficientNet: Lightweight Architectures for GAN-Based Face Forensics https://arxiv.org/abs/2609.01749 https://arxiv.org/pdf/2609.01749 https://arxiv.org/html/2609.01749
September 3, 2026 at 6:39 AM
Learning Radio Astronomical Representations with LeJEPA and Very Small Models. Erica Lastufka et. al. https://arxiv.org/abs/2608.30594
September 1, 2026 at 6:27 AM
Signal or noise? > AI, Vol. 7, Pages 336: X-Net: Hybrid
Transformer–CNN Framework with Multi-Aspect Attention
for Precise Colon Polyp Segmentation >> Comment below! #healthtech #IoT #industry40 #AI #mhealth
AI, Vol. 7, Pages 336: X-Net: Hybrid Transformer–CNN Framework with Multi-Aspect Attention for Precise Colon Polyp Segmentation
Accurate localization of colon polyps plays a crucial role in the early diagnosis of colon cancer. In this paper, we propose a novel network architecture, X-Net, which integrates the Pyramid Vision Transformer (PVT) and EfficientNet-B0. We selected EfficientNet-B0 due to its optimized design and computational efficiency, enabling it to effectively capture fine-grained local features. Meanwhile, we use PVT V2-b1 (Pb1) because of its hierarchical structure and self-attention mechanism, which enhance the model’s ability to preserve multi-scale features. By concatenating these two networks in part of the decoder, the proposed model effectively leverages both local and global features. Additionally, the X-Net architecture introduces three blocks: Shape-Aware Enhancement Block (SAEB), Multi-Scale Feature Block (MSFB), and Multi-Aspect Attention Block (MAAB). The SAEB utilizes differences in initial-layer features extracted from the encoder to filter noise in polyp images. The MSFB extracts and enhances multi-scale features from the backbone, effectively integrating semantic information, which improves polyp segmentation accuracy and enhances robustness against variations in polyp size and shape. Finally, the MAAB employs high-level features from the outputs of the MSFB and SAEB to attend to the foreground, background, and boundaries. To address the class imbalance problem, we propose a hybrid loss function combining Tversky Loss and Binary Cross-Entropy Loss. For evaluation, we trained the proposed network on four polyp datasets: Kvasir-SEG, CVC-ClinicDB, CVC-T, and ETIS. The results demonstrate that the model performs well across different datasets with varying image quality and characteristics.
dlvr.it
August 30, 2026 at 1:05 PM
튜링의 논문부터 ChatGPT, Sora, DeepSeek-R1까지 – AI의 과거·현재·미래를 한 번에 정리한 역사 끝판왕 가이드. DBN부터 EfficientNet까지 빠진 모델 없이 완벽 정리! #AI역사­사 #AlphaGo #ChatGPT #CNN #EfficientNet #ResNet #Transformer #딥러닝러닝 #생성형AI±형AI #인공지능³µ지능 doyouknow.kr/846/ai-histo...
튜링에서 DeepSeek까지, AI 80년의 모든 것 – 겨울 2번 버티고 폭발한 진짜 역사
AI의 역사는 '과장된 약속 → 혹독한 겨울 → 하드웨어·데이터·알고리즘 삼박자의 혁신'이 반복된 이야기예요. 지금의 Ch
doyouknow.kr
December 9, 2025 at 4:36 AM
Vision Transformer(ViT) 완벽 분석! 이미지를 16×16 패치로 분할, Self-Attention으로 전역 패턴 학습. ViT-H/14 ImageNet 88.55% 달성, ResNet보다 4배 효율적! CNN vs ViT 성능 비교, Inductive Bias 차이, CLIP 텍스트-이미지 연결, Segment Anything(SAM) 범용 분할까지 완벽 가이드.

#CLIP #CNN #EfficientNet #InductiveBias #ResNet #SAM
doyouknow.kr/602/vision-t...
December 3, 2025 at 2:51 AM
I used EfficientNet (lmz/candle-efficientnet from huggingface), there's also an option to use convnext but it runs like unusably slow for me because i can't use gpu accel
May 24, 2026 at 9:21 PM
Hirose et al. OmniVLA: An Omni-Modal Vision-Language-Action Model for Robot Navigation
arxiv.org/pdf/2509.19480

Navigation VLA (as opposed to robotic manipulation) + conditioning on language, ego-centric & satellite images, GPS-ish poses
2 flavours: 7B Llama + 50M edge version (EfficientNet, CLIP)
arxiv.org
September 25, 2025 at 8:44 AM
So what's the point of these processes? Why would V1 have them if they don't help explain V1 activity?

It seems like all you need is an adversarially trained ResNet, ResNeXt, or EfficientNet... and the rest will follow.

Should #ML/#AI even care about these biological details?
November 9, 2024 at 8:57 PM
Leveraging Vision-Language Models for Resource Constrained Settings

Anna Bair, Madan Ravi Ganesh, Devin Willmott, J Zico Kolter

Action editor: Massimiliano Mancini

https://openreview.net/forum?id=cYOKSg60jC

#efficientnet #leveraging #trained
August 10, 2026 at 12:20 PM
And these models are just classifiers (ResNet / EfficientNet) and importantly are NOT Generative Adversarial Networks, which would have a component that would generate imagery for obvious reasons. Some approaches train classifiers on commercial adult pornography for human review
February 11, 2025 at 6:43 PM
Electronics, Vol. 14, Pages 1084: Optimized Real-Time Decision Making with EfficientNet in Digital Twin-Based Vehicular Networks
Real-time decision-making is vital in vehicular ad hoc networks (VANETs). It is essential to improve road safety and ensure traffic efficiency and flow. Integrating digital twins within VANET (DT-VANET) creates virtual replicas of physical vehicles, allowing in-depth analysis and effective decision-making. Many vehicular ad hoc network applications now use convolutional neural networks (CNNs). However, the growing model size and latency make implementing them in real-time systems challenging, and most previous studies focusing on using CNNs still face significant challenges. Some effective models with sustainable performances have recently been proposed. One of the most advanced models among them is EfficientNet. One may consider it a family of network models with significantly fewer parameters and computational costs. This paper proposes EfficientNet-based optimized real-time decision-making in the DT-VANET architecture. This paper investigates the performance of EfficientNet in digital-based vehicular ad hoc networks. Extensive experiments have proved that EfficientNet outperforms CNN models (ResNet50, VGG16) in accuracy, latency, computational efficiency, and convergence time, which proves its effectiveness in real-time applications of DT-VANET.
www.mdpi.com
March 9, 2025 at 10:47 AM
Sensors, Vol. 25, Pages 1478: Application of Multiple Deep Learning Architectures for Emotion Classification Based on Facial Expressions
Facial expression recognition (FER) is essential for discerning human emotions and is applied extensively in big data analytics, healthcare, security, and user experience enhancement. This study presents a comprehensive evaluation of ten state-of-the-art deep learning models—VGG16, VGG19, ResNet50, ResNet101, DenseNet, GoogLeNet V1, MobileNet V1, EfficientNet V2, ShuffleNet V2, and RepVGG—on the task of facial expression recognition using the FER2013 dataset. Key performance metrics, including test accuracy, training time, and weight file size, were analyzed to assess the learning efficiency, generalization capabilities, and architectural innovations of each model. EfficientNet V2 and ResNet50 emerged as top performers, achieving high accuracy and stable convergence using compound scaling and residual connections, enabling them to capture complex emotional features with minimal overfitting. DenseNet, GoogLeNet V1, and RepVGG also demonstrated strong performance, leveraging dense connectivity, inception modules, and re-parameterization techniques, though they exhibited slower initial convergence. In contrast, lightweight models such as MobileNet V1 and ShuffleNet V2, while excelling in computational efficiency, faced limitations in accuracy, particularly in challenging emotion categories like “fear” and “disgust”. The results highlight the critical trade-offs between computational efficiency and predictive accuracy, emphasizing the importance of selecting appropriate architecture based on application-specific requirements. This research contributes to ongoing advancements in deep learning, particularly in domains such as facial expression recognition, where capturing subtle and complex patterns is essential for high-performance outcomes.
www.mdpi.com
February 27, 2025 at 6:03 PM
この研究を深く理解したい方には、Googleの「EfficientNet」という論文もおすすめです。モデルの幅・深さ・解像度をいかにバランスよく調整するかという、実用的な最適化の視点が学べます。

また、脳とAIの構造的な共通点に興味があるなら、ジェフ・ホーキンス著『脳は世界をどう見ているのか』も必読の一冊。知能の本質に迫るヒントが満載です。

「限られたリソースで最大の成果を出す」というこの考え方、あなたの今の活動にも取り入れられそうな部分はありますか?ぜひコメントで教えてください!💡
May 26, 2026 at 11:32 PM
recordings from this dataset, transformed into normalized mel-spectrograms. For classification, the EfficientNet-B0 architecture was used, enhanced with a Squeeze-and-Excitation attention mechanism. The developed model achieved accuracy rates of [4/6 of https://arxiv.org/abs/2503.23470v1]
April 1, 2025 at 6:01 AM
Sophie Zeng, Sean Kalaycioglu, Collin Hong, Haipeng Xie: Interpretable Image-Level Acne Severity Grading via EfficientNet-B0 Transfer Learning and Grad-CAM https://arxiv.org/abs/2607.26461 https://arxiv.org/pdf/2607.26461 https://arxiv.org/html/2607.26461
July 30, 2026 at 6:40 AM