SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1911.11907 · 2019

GhostNet: More Features from Cheap Operations

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 23 functions out of this paper's own repositories and ran 19 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
0jason000/S-GhostNet pwc_unofficial 7 of 7
ozora-ogino/efficient_backbones pwc_unofficial 5 of 7
0jason000/ghostnet pwc_unofficial 2 of 3
yangyucheng000/ghostnet pwc_unofficial 2 of 2
copy not recorded — 2 of 2
KopiSoftware/Ghost_ResNet56 reimplementation 1 of 1
iamhankai/ghostnet.pytorch reimplementation 0 of 1
FunctionStatusWhere it lives
conv1x1 Ran ozora-ogino/efficient_backbones/efficient_backbones/g_ghost_regnet.py
code served (permissive licence) · get_code("d9def42110729a85")
conv1x1 Ran ozora-ogino/efficient_backbones/efficient_backbones/resnet.py
code served (permissive licence) · get_code("6f0d1921e77f5311")
conv3x3 Ran ozora-ogino/efficient_backbones/efficient_backbones/g_ghost_regnet.py
code served (permissive licence) · get_code("160bb14bd76201b4")
conv3x3 Ran ozora-ogino/efficient_backbones/efficient_backbones/resnet.py
code served (permissive licence) · get_code("7e71d6f8569b6b34")
cutout Ran KopiSoftware/Ghost_ResNet56/train_Ghost_ResNet56.py
pointer only (licence: GPL-3.0) · get_code("ce3aa75b714c404e")
float_parameter Ran 0jason000/S-GhostNet/src/autoaug.py
code served (permissive licence) · get_code("40d3c39a403ae1f7")
gen_cfgs_1x Ran 0jason000/S-GhostNet/src/big_net.py
code served (permissive licence) · get_code("7022c26a18d085ec")
gen_cfgs_large Ran 0jason000/S-GhostNet/src/big_net.py
code served (permissive licence) · get_code("c83dff446ca0238d")
get_lr Ran 0jason000/S-GhostNet/src/utils.py
code served (permissive licence) · get_code("9dfb4495e7fb2efb")
get_lr Ran yangyucheng000/ghostnet/src/lr_generator.py
code served (permissive licence) · get_code("78ea6999b2c97419")
get_lr_tinynet_c Ran 0jason000/S-GhostNet/src/utils.py
code served (permissive licence) · get_code("e1258604a87c9da4")
get_top5_acc Ran yangyucheng000/ghostnet/postprocess.py
code served (permissive licence) · get_code("fca5ffc375803fa4")
hard_sigmoid Ran ozora-ogino/efficient_backbones/efficient_backbones/ghostnet.py
code served (permissive licence) · get_code("ca43d9eb44b430e5")
init_group_params Ran 0jason000/ghostnet/src/optim.py
code served (permissive licence) · get_code("698131638ff8d4ad")
int_parameter Ran 0jason000/S-GhostNet/src/autoaug.py
code served (permissive licence) · get_code("e5a370ee6856bd8d")
is_model Ran 0jason000/ghostnet/src/model.py
code served (permissive licence) · get_code("b85a6569fd4f3890")
round_filters Ran this paper's copy was not recorded; identical code first harvested from seermer/TensorFlow2-EfficientNetV2
pointer only · get_code("e60387e9529f63ff")
round_repeats Ran this paper's copy was not recorded; identical code first harvested from seermer/TensorFlow2-EfficientNetV2
pointer only · get_code("1a9d7e65e0a9fbc9")
str2bool Ran 0jason000/S-GhostNet/src/utils.py
code served (permissive licence) · get_code("2dedef59d5e447a2")
conv_1x1_bn Not yet run ozora-ogino/efficient_backbones/efficient_backbones/efficientnet_v2.py
code served (permissive licence) · get_code("f76badd0c6766f8a")
conv_3x3_bn Not yet run ozora-ogino/efficient_backbones/efficient_backbones/efficientnet_v2.py
code served (permissive licence) · get_code("3d6f0eda22ace06f")
depthwise_conv Not yet run iamhankai/ghostnet.pytorch/ghost_net.py
pointer only (licence: NONE) · get_code("680d03183fe725d4")
model_entrypoint Not yet run 0jason000/ghostnet/src/model.py
code served (permissive licence) · get_code("ac95eb88d9b993fd")

Repositories linked to this paper

More are linked than shown here.

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Deploying convolutional neural networks (CNNs) on embedded devices is difficult due to the limited memory and computation resources. The redundancy in feature maps is an important characteristic of those successful CNNs, but has rarely been investigated in neural architecture design. This paper proposes a novel Ghost module to generate more feature maps from cheap operations. Based on a set of intrinsic feature maps, we apply a series of linear transformations with cheap cost to generate many ghost feature maps that could fully reveal information underlying intrinsic features. The proposed Ghost module can be taken as a plug-and-play component to upgrade existing convolutional neural networks. Ghost bottlenecks are designed to stack Ghost modules, and then the lightweight GhostNet can be easily established. Experiments conducted on benchmarks demonstrate that the proposed Ghost module is an impressive alternative of convolution layers in baseline models, and our GhostNet can achieve higher recognition performance (e.g. $75.7\%$ top-1 accuracy) than MobileNetV3 with similar computational cost on the ImageNet ILSVRC-2012 classification dataset. Code is available at https://github.com/huawei-noah/ghostnet

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("1911.11907")
get_code_for_paper("1911.11907")
have("1911.11907")

Connect an agent — have() is free.