Paddle-Image-Models

A PaddlePaddle version image model zoo.

Model Zoo
CNN	Transformer	MLP
DLA ReXNet RedNet RepVGG HardNet CondenseNet V2	PiT PvT TNT DeiT CaiT CoaT LeViT LV ViT T2T ViT Swin Transformer	MLP-Mixer

Install Package

Install by pip：
```
$ pip install ppim
```
Install by wheel package：【Releases Packages】

Usage

Quick Start

import paddle
from ppim import rednet_26

# Load the model with PPIM wheel package
model, val_transforms = rednet_26(pretrained=True, return_transforms=True)

# Load the model with paddle.hub API
# paddlepaddle >= 2.1.0
'''
model, val_transforms = paddle.hub.load(
    'AgentMaker/Paddle-Image-Models:dev', 
    'rednet_26', 
    source='github', 
    force_reload=False, 
    pretrained=True, 
    return_transforms=True
)
'''

# Model summary 
paddle.summary(model, input_size=(1, 3, 224, 224))

# Random a input
x = paddle.randn(shape=(1, 3, 224, 224))

# Model forword
out = model(x)

Classification（PaddleHapi）

import paddle
import paddle.nn as nn
import paddle.vision.transforms as T
from paddle.vision import Cifar100

from ppim import rexnet_1_0

# Load the model
model, val_transforms = rexnet_1_0(pretrained=True, return_transforms=True, class_dim=100)

# Use the PaddleHapi Model
model = paddle.Model(model)

# Set the optimizer
opt = paddle.optimizer.Adam(learning_rate=0.001, parameters=model.parameters())

# Set the loss function
loss = nn.CrossEntropyLoss()

# Set the evaluate metric
metric = paddle.metric.Accuracy(topk=(1, 5))

# Prepare the model 
model.prepare(optimizer=opt, loss=loss, metrics=metric)

# Set the data preprocess
train_transforms = T.Compose([
    T.Resize(256, interpolation='bicubic'),
    T.RandomCrop(224),
    T.ToTensor(),
    T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

# Load the Cifar100 dataset
train_dataset = Cifar100(mode='train', transform=train_transforms, backend='pil')
val_dataset = Cifar100(mode='test',  transform=val_transforms, backend='pil')

# Finetune the model 
model.fit(
    train_data=train_dataset, 
    eval_data=val_dataset, 
    batch_size=256, 
    epochs=2, 
    eval_freq=1, 
    log_freq=1, 
    save_dir='save_models', 
    save_freq=1, 
    verbose=1, 
    drop_last=False, 
    shuffle=True,
    num_workers=0
)

Contact us

Email : agentmaker@163.com
QQ Group : 1005109853