Common CNN Architectures

~10 min · resnet, efficientnet, mobilenet, convnext

Level 0Level 0

0 XP0/78 lessons0/17 achievements

0/100 XP to next level100 XP to go0% complete

The architectures you should recognize

Knowing the canonical architectures helps you pick the right backbone and read other people's code without confusion.

Architecture	Year	Key idea	Use when
ResNet	2015	Skip connections enable very deep nets	Solid default; 50/101/152 layers
MobileNetV2/V3	2018-19	Depthwise separable conv	Mobile / edge deployment
EfficientNetV2	2021	Compound scaling of depth/width/resolution	State-of-art accuracy/cost balance
ConvNeXt	2022	CNN tuned with transformer training tricks	Pure-CNN baseline competitive with ViT
ViT (Vision Transformer)	2020	Self-attention over image patches	Large datasets, very large models

All of these are available pretrained via keras.applications or keras_cv.models. For most projects, EfficientNetV2-S is a strong default — best accuracy-per-FLOP tradeoff for mid-range hardware.

Code

Pretrained 모델 즉시 사용·python

import keras

# ImageNet-pretrained classifiers (full top layer)
resnet = keras.applications.ResNet50(weights='imagenet')
mobilenet = keras.applications.MobileNetV3Small(weights='imagenet')
efficientnet = keras.applications.EfficientNetV2S(weights='imagenet')

# Backbone only (no classifier head) — for transfer learning
backbone = keras.applications.EfficientNetV2S(
    weights='imagenet',
    include_top=False,
    input_shape=(224, 224, 3),
)

# Inference on a single image
import tensorflow as tf
img = tf.keras.utils.load_img("cat.jpg", target_size=(224, 224))
img_array = tf.keras.utils.img_to_array(img)
img_batch = tf.expand_dims(img_array, 0)
preds = resnet.predict(img_batch)
print(keras.applications.resnet50.decode_predictions(preds, top=5))

External links

keras.applications

Progress

Progress is local-only — sign in to sync across devices.

← PreviousTransfer Learning — Feature Extraction vs Fine-Tuning Next →Grad-CAM — Seeing What CNNs See

Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)

No comments yet — be the first.