AI Rank
EN
← 开源
ethereon

caffe-tensorflow

Caffe models in TensorFlow

InfrastructureModel optimizationPython
在 GitHub 打开
增长势头
+024 小时新增 Star0.0%
2.79k
Star
1.01k
Fork
+0
本周
5
贡献者
创建于 2015-11-10 · 更新于 2026-10-01 · 今日第 8370 名
主要开发者
ethereonraingomarkusnagel

ethereon/caffe-tensorflow 排名第 8,370，有 2,786 个星标。

同类别比较

ethereon/caffe-tensorflow 排名第 8,370，此处有 8 个同类条目；FasterDecoding/Medusa 排名第 8,378。

同类别比较
名称排名指标
whylabs/whylogs8,336
语言:
Jupyter Notebook
星标数:
2,834
分叉数:
144
简介:
An open-source data logging library for machine learning models and data pipelines. 📚 Provides visibility into data quality & model performance over time. 🛡️ Supports privacy-preserving data collection, ensuring safety & robustness. 📈
leptonai/leptonai8,343
语言:
Python
星标数:
2,824
分叉数:
198
简介:
A Pythonic framework to simplify AI service building
rom1504/clip-retrieval8,360
语言:
Jupyter Notebook
星标数:
2,800
分叉数:
236
简介:
Easily compute clip embeddings and build a clip retrieval system with them
facebookresearch/hiplot8,361
语言:
TypeScript
星标数:
2,797
分叉数:
148
简介:
HiPlot makes understanding high dimensional data easy
FasterDecoding/Medusa8,378
语言:
Jupyter Notebook
星标数:
2,775
分叉数:
206
简介:
Medusa: Simple Framework for Accelerating LLM Generation with Multiple Decoding Heads
apple-aiml-research/ml-ane-transformers8,392
语言:
Python
星标数:
2,747
分叉数:
93
简介:
Reference implementation of the Transformer architecture optimized for Apple Neural Engine (ANE)
agentrhq/webcmd8,409
语言:
TypeScript
星标数:
2,713
分叉数:
1,547
简介:
Self-learning agent browser
intel/neural-compressor8,412
语言:
Python
星标数:
2,712
分叉数:
329
简介:
SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime
关键参数与规格
关键参数与规格
参数值
语言Python
类别Infrastructure
子类别Model optimization
星标数2,786
分叉数1,012
贡献者5
创建日期2015-11-10
更新日期2026-10-01
1 天星标数0
7 天星标数0
排名8,370
上次排名8,422
排名变化52
7 天星标排名11,962
上次 7 天星标排名12,013
7 天星标排名变化51
总星标排名4,847
上次总星标排名4,843
总星标排名变化-4
国家USA
README 摘录

Caffe to TensorFlow Convert Caffe models to TensorFlow. Usage Run to convert an existing Caffe model to TensorFlow. Make sure you're using the latest Caffe format (see the notes section for more info). A data file (in NumPy's native format) containing the model's learned parameters. A Python class that constructs the model's graph. Examples See the examples folder for more details. Verification The following converted models have been verified on the ILSVRC2012 validation set using validate. 84% | Notes Only the

new Caffe model format is supported. If you have an old model, use the and tools that ship with Caffe to upgrade them first. Also make sure you're using a fairly recent version of Caffe. It appears that Caffe and TensorFlow cannot be concurrently invoked (CUDA conflicts - even with ). This makes it a two-stage process: first extract the parameters with , then import it into TensorFlow. If PyCaffe is found in your , and the environment variable is set, it will

be used. However, the fallback uses the pure Python-based implementation of protobuf, which is astoundingly slow (~1. 5 minutes to parse the VGG16 parameters). The experimental CPP protobuf backend doesn't particularly help here, since it runs into the file size limit (Caffe gets around this by overriding this limit in C++). A cleaner solution here would be to implement the loader as a C++ module. Only a subset of Caffe layers and accompanying parameters are currently supported. Not all

Caffe models can be converted to TensorFlow. For instance, Caffe supports arbitrary padding whereas TensorFlow's support is currently restricted to and . The border values are handled differently by Caffe and TensorFlow. However, these don't appear to affect things too much. Image rescaling can affect the ILSVRC2012 top 5 accuracy listed above slightly. VGG16 expects isotropic rescaling (anisotropic reduces accuracy to 88. 45%) whereas BVLC's implementation of GoogLeNet expects anisotropic (isotropic reduces accuracy to 87. The support class has no internal

dependencies. It can be safely extracted and deployed without the rest of this library. The ResNet model uses 1x1 convolutions with a stride of 2. This is currently only supported in the master branch of TensorFlow (the latest release at time of writing being v0.

常见问题

记录了多少星标和分叉？

ethereon/caffe-tensorflow 有 2,786 个星标和 1,012 个分叉。

仓库排名是多少？

ethereon/caffe-tensorflow 排名第 8,370。

记录的语言和类别是什么？

ethereon/caffe-tensorflow 有 2,786 个星标，语言为 Python，类别为 Infrastructure。

哪个仓库排名最接近？

ethereon/caffe-tensorflow 与 FasterDecoding/Medusa 比较，后者排名第 8,378，有 2,775 个星标。

README

Caffe to TensorFlow

Convert Caffe models to TensorFlow.

Usage

Run convert.py to convert an existing Caffe model to TensorFlow.

Make sure you're using the latest Caffe format (see the notes section for more info).

The output consists of two files:

  1. A data file (in NumPy's native format) containing the model's learned parameters.
  2. A Python class that constructs the model's graph.

Examples

See the examples folder for more details.

Verification

The following converted models have been verified on the ILSVRC2012 validation set using validate.py.

Model Top 5 Accuracy
ResNet 152 92.92%
ResNet 101 92.63%
ResNet 50 92.02%
VGG 16 89.88%
GoogLeNet 89.06%
Network in Network 81.21%
CaffeNet 79.93%
AlexNet 79.84%

Notes

  • Only the new Caffe model format is supported. If you have an old model, use the upgrade_net_proto_text and upgrade_net_proto_binary tools that ship with Caffe to upgrade them first. Also make sure you're using a fairly recent version of Caffe.

  • It appears that Caffe and TensorFlow cannot be concurrently invoked (CUDA conflicts - even with set_mode_cpu). This makes it a two-stage process: first extract the parameters with convert.py, then import it into TensorFlow.

  • Caffe is not strictly required. If PyCaffe is found in your PYTHONPATH, and the USE_PYCAFFE environment variable is set, it will be used. Otherwise, a fallback will be used. However, the fallback uses the pure Python-based implementation of protobuf, which is astoundingly slow (~1.5 minutes to parse the VGG16 parameters). The experimental CPP protobuf backend doesn't particularly help here, since it runs into the file size limit (Caffe gets around this by overriding this limit in C++). A cleaner solution here would be to implement the loader as a C++ module.

  • Only a subset of Caffe layers and accompanying parameters are currently supported.

  • Not all Caffe models can be converted to TensorFlow. For instance, Caffe supports arbitrary padding whereas TensorFlow's support is currently restricted to SAME and VALID.

  • The border values are handled differently by Caffe and TensorFlow. However, these don't appear to affect things too much.

  • Image rescaling can affect the ILSVRC2012 top 5 accuracy listed above slightly. VGG16 expects isotropic rescaling (anisotropic reduces accuracy to 88.45%) whereas BVLC's implementation of GoogLeNet expects anisotropic (isotropic reduces accuracy to 87.7%).

  • The support class kaffe.tensorflow.Network has no internal dependencies. It can be safely extracted and deployed without the rest of this library.

  • The ResNet model uses 1x1 convolutions with a stride of 2. This is currently only supported in the master branch of TensorFlow (the latest release at time of writing being v0.8.0, which does not support it).

排名相邻
同类项目