From 7e83d3175ff2e0849738b8b5819487e6fabfedbe Mon Sep 17 00:00:00 2001 From: liaoxingyu Date: Mon, 18 Jan 2021 11:44:55 +0800 Subject: [PATCH] update README.md Summary: add information about fastreid V1.0 --- README.md | 11 +- fastreid/modeling/backbones/regnet/config.py | 28 +---- tools/deploy/Caffe/layer_param.py | 109 +++++++++---------- tools/deploy/Caffe/net.py | 3 +- 4 files changed, 61 insertions(+), 90 deletions(-) diff --git a/README.md b/README.md index e4b8a0832..823b815cd 100644 --- a/README.md +++ b/README.md @@ -4,13 +4,14 @@ FastReID is a research platform that implements state-of-the-art re-identificati ## What's New -- [Oct 2020] Added the [Hyper-Parameter Optimization](https://github.com/JDAI-CV/fast-reid/tree/master/projects/HPOReID) based on fastreid. See `projects/HPOReID`. -- [Sep 2020] Added the [person attribute recognition](https://github.com/JDAI-CV/fast-reid/tree/master/projects/attribute_recognition) based on fastreid. See `projects/attribute_recognition`. -- [Sep 2020] Automatic Mixed Precision training is supported with pytorch1.6 built-in `torch.cuda.amp`. Set `cfg.SOLVER.AMP_ENABLED=True` to switch it on. -- [Aug 2020] [Model Distillation](https://github.com/JDAI-CV/fast-reid/tree/master/projects/DistillReID) is supported, thanks for [guan'an wang](https://github.com/wangguanan)'s contribution. +- [Jan 2021] FastReID V1.0 has been released！🎉 +Support many tasks beyond reid, such image retrieval and face recognition. See [projects](https://github.com/JDAI-CV/fast-reid/tree/master/projects). +- [Oct 2020] Added the [Hyper-Parameter Optimization](https://github.com/JDAI-CV/fast-reid/tree/master/projects/FastTune) based on fastreid. See `projects/FastTune`. +- [Sep 2020] Added the [person attribute recognition](https://github.com/JDAI-CV/fast-reid/tree/master/projects/FastAttr) based on fastreid. See `projects/FastAttr`. +- [Sep 2020] Automatic Mixed Precision training is supported with `apex`. Set `cfg.SOLVER.FP16_ENABLED=True` to switch it on. +- [Aug 2020] [Model Distillation](https://github.com/JDAI-CV/fast-reid/tree/master/projects/FastDistill) is supported, thanks for [guan'an wang](https://github.com/wangguanan)'s contribution. - [Aug 2020] ONNX/TensorRT converter is supported. - [Jul 2020] Distributed training with multiple GPUs, it trains much faster. -- [Jul 2020] `MAX_ITER` in config means `epoch`, it will auto scale to maximum iterations. - Includes more features such as circle loss, abundant visualization methods and evaluation metrics, SoTA results on conventional, cross-domain, partial and vehicle re-id, testing on multi-datasets simultaneously, etc. - Can be used as a library to support [different projects](https://github.com/JDAI-CV/fast-reid/tree/master/projects) on top of it. We'll open source more research projects in this way. - Remove [ignite](https://github.com/pytorch/ignite)(a high-level library) dependency and powered by [PyTorch](https://pytorch.org/). diff --git a/fastreid/modeling/backbones/regnet/config.py b/fastreid/modeling/backbones/regnet/config.py index 96f764930..4496480fd 100644 --- a/fastreid/modeling/backbones/regnet/config.py +++ b/fastreid/modeling/backbones/regnet/config.py @@ -13,7 +13,6 @@ from yacs.config import CfgNode as CfgNode - # Global config object _C = CfgNode() @@ -21,7 +20,6 @@ # from core.config import cfg cfg = _C - # ------------------------------------------------------------------------------------ # # Model options # ------------------------------------------------------------------------------------ # @@ -39,7 +37,6 @@ # Loss function (see pycls/models/loss.py for options) _C.MODEL.LOSS_FUN = "cross_entropy" - # ------------------------------------------------------------------------------------ # # ResNet options # ------------------------------------------------------------------------------------ # @@ -57,7 +54,6 @@ # Apply stride to 1x1 conv (True -> MSRA; False -> fb.torch) _C.RESNET.STRIDE_1X1 = True - # ------------------------------------------------------------------------------------ # # AnyNet options # ------------------------------------------------------------------------------------ # @@ -93,7 +89,6 @@ # SE ratio _C.ANYNET.SE_R = 0.25 - # ------------------------------------------------------------------------------------ # # RegNet options # ------------------------------------------------------------------------------------ # @@ -133,7 +128,6 @@ # Bottleneck multiplier (bm = 1 / b from the paper) _C.REGNET.BOT_MUL = 1.0 - # ------------------------------------------------------------------------------------ # # EfficientNet options # ------------------------------------------------------------------------------------ # @@ -169,7 +163,6 @@ # Dropout ratio _C.EN.DROPOUT_RATIO = 0.0 - # ------------------------------------------------------------------------------------ # # Batch norm options # ------------------------------------------------------------------------------------ # @@ -192,7 +185,6 @@ _C.BN.USE_CUSTOM_WEIGHT_DECAY = False _C.BN.CUSTOM_WEIGHT_DECAY = 0.0 - # ------------------------------------------------------------------------------------ # # Optimizer options # ------------------------------------------------------------------------------------ # @@ -234,7 +226,6 @@ # Gradually warm up the OPTIM.BASE_LR over this number of epochs _C.OPTIM.WARMUP_EPOCHS = 0 - # ------------------------------------------------------------------------------------ # # Training options # ------------------------------------------------------------------------------------ # @@ -262,7 +253,6 @@ # Weights to start training from _C.TRAIN.WEIGHTS = "" - # ------------------------------------------------------------------------------------ # # Testing options # ------------------------------------------------------------------------------------ # @@ -281,7 +271,6 @@ # Weights to use for testing _C.TEST.WEIGHTS = "" - # ------------------------------------------------------------------------------------ # # Common train/test data loader options # ------------------------------------------------------------------------------------ # @@ -293,7 +282,6 @@ # Load data to pinned host memory _C.DATA_LOADER.PIN_MEMORY = True - # ------------------------------------------------------------------------------------ # # Memory options # ------------------------------------------------------------------------------------ # @@ -302,7 +290,6 @@ # Perform ReLU inplace _C.MEM.RELU_INPLACE = True - # ------------------------------------------------------------------------------------ # # CUDNN options # ------------------------------------------------------------------------------------ # @@ -313,7 +300,6 @@ # in overall speedups when variable size inputs are used (e.g. COCO training) _C.CUDNN.BENCHMARK = True - # ------------------------------------------------------------------------------------ # # Precise timing options # ------------------------------------------------------------------------------------ # @@ -325,7 +311,6 @@ # Number of iterations to compute avg time _C.PREC_TIME.NUM_ITER = 30 - # ------------------------------------------------------------------------------------ # # Misc options # ------------------------------------------------------------------------------------ # @@ -359,7 +344,6 @@ # Models weights referred to by URL are downloaded to this local cache _C.DOWNLOAD_CACHE = "/tmp/pycls-download-cache" - # ------------------------------------------------------------------------------------ # # Deprecated keys # ------------------------------------------------------------------------------------ # @@ -369,7 +353,7 @@ _C.register_deprecated_key("PORT") -def assert_and_infer_cfg(cache_urls=True): +def assert_and_infer_cfg(): """Checks config values invariants.""" err_str = "The first lr step must start at 0" assert not _C.OPTIM.STEPS or _C.OPTIM.STEPS[0] == 0, err_str @@ -382,14 +366,6 @@ def assert_and_infer_cfg(cache_urls=True): assert _C.TEST.BATCH_SIZE % _C.NUM_GPUS == 0, err_str err_str = "Log destination '{}' not supported" assert _C.LOG_DEST in ["stdout", "file"], err_str.format(_C.LOG_DEST) - if cache_urls: - cache_cfg_urls() - - -def cache_cfg_urls(): - """Download URLs in config, cache them, and rewrite cfg to use cached file.""" - _C.TRAIN.WEIGHTS = cache_url(_C.TRAIN.WEIGHTS, _C.DOWNLOAD_CACHE) - _C.TEST.WEIGHTS = cache_url(_C.TEST.WEIGHTS, _C.DOWNLOAD_CACHE) def dump_cfg(): @@ -417,4 +393,4 @@ def load_cfg_fom_args(description="Config file options."): sys.exit(1) args = parser.parse_args() _C.merge_from_file(args.cfg_file) - _C.merge_from_list(args.opts) \ No newline at end of file + _C.merge_from_list(args.opts) diff --git a/tools/deploy/Caffe/layer_param.py b/tools/deploy/Caffe/layer_param.py index 1e13a2426..05d1994a0 100644 --- a/tools/deploy/Caffe/layer_param.py +++ b/tools/deploy/Caffe/layer_param.py @@ -1,11 +1,12 @@ from __future__ import absolute_import + from . import caffe_pb2 as pb -import numpy as np -def pair_process(item,strict_one=True): - if hasattr(item,'__iter__'): + +def pair_process(item, strict_one=True): + if hasattr(item, '__iter__'): for i in item: - if i!=item[0]: + if i != item[0]: if strict_one: raise ValueError("number in item {} must be the same".format(item)) else: @@ -13,26 +14,28 @@ def pair_process(item,strict_one=True): return item[0] return item + def pair_reduce(item): - if hasattr(item,'__iter__'): + if hasattr(item, '__iter__'): for i in item: - if i!=item[0]: + if i != item[0]: return item return [item[0]] return [item] + class Layer_param(): - def __init__(self,name='',type='',top=(),bottom=()): - self.param=pb.LayerParameter() - self.name=self.param.name=name - self.type=self.param.type=type + def __init__(self, name='', type='', top=(), bottom=()): + self.param = pb.LayerParameter() + self.name = self.param.name = name + self.type = self.param.type = type - self.top=self.param.top + self.top = self.param.top self.top.extend(top) - self.bottom=self.param.bottom + self.bottom = self.param.bottom self.bottom.extend(bottom) - def fc_param(self, num_output, weight_filler='xavier', bias_filler='constant',has_bias=True): + def fc_param(self, num_output, weight_filler='xavier', bias_filler='constant', has_bias=True): if self.type != 'InnerProduct': raise TypeError('the layer type must be InnerProduct if you want set fc param') fc_param = pb.InnerProductParameter() @@ -45,7 +48,7 @@ def fc_param(self, num_output, weight_filler='xavier', bias_filler='constant',ha def conv_param(self, num_output, kernel_size, stride=(1), pad=(0,), weight_filler_type='xavier', bias_filler_type='constant', - bias_term=True, dilation=None,groups=None): + bias_term=True, dilation=None, groups=None): """ add a conv_param layer if you spec the layer type "Convolution" Args: @@ -56,80 +59,69 @@ def conv_param(self, num_output, kernel_size, stride=(1), pad=(0,), bias_filler_type: the bias filler type Returns: """ - if self.type not in ['Convolution','Deconvolution']: + if self.type not in ['Convolution', 'Deconvolution']: raise TypeError('the layer type must be Convolution or Deconvolution if you want set conv param') - conv_param=pb.ConvolutionParameter() - conv_param.num_output=num_output + conv_param = pb.ConvolutionParameter() + conv_param.num_output = num_output conv_param.kernel_size.extend(pair_reduce(kernel_size)) conv_param.stride.extend(pair_reduce(stride)) conv_param.pad.extend(pair_reduce(pad)) - conv_param.bias_term=bias_term - conv_param.weight_filler.type=weight_filler_type + conv_param.bias_term = bias_term + conv_param.weight_filler.type = weight_filler_type if bias_term: conv_param.bias_filler.type = bias_filler_type if dilation: conv_param.dilation.extend(pair_reduce(dilation)) if groups: - conv_param.group=groups + conv_param.group = groups self.param.convolution_param.CopyFrom(conv_param) - def pool_param(self,type='MAX',kernel_size=2,stride=2,pad=None, ceil_mode = False): - pool_param=pb.PoolingParameter() - pool_param.pool=pool_param.PoolMethod.Value(type) - pool_param.kernel_size=pair_process(kernel_size) - pool_param.stride=pair_process(stride) - pool_param.ceil_mode=ceil_mode + def pool_param(self, type='MAX', kernel_size=2, stride=2, pad=None, ceil_mode=False): + pool_param = pb.PoolingParameter() + pool_param.pool = pool_param.PoolMethod.Value(type) + pool_param.kernel_size = pair_process(kernel_size) + pool_param.stride = pair_process(stride) + pool_param.ceil_mode = ceil_mode if pad: - if isinstance(pad,tuple): + if isinstance(pad, tuple): pool_param.pad_h = pad[0] pool_param.pad_w = pad[1] else: - pool_param.pad=pad + pool_param.pad = pad self.param.pooling_param.CopyFrom(pool_param) - def batch_norm_param(self,use_global_stats=0,moving_average_fraction=None,eps=None): - bn_param=pb.BatchNormParameter() - bn_param.use_global_stats=use_global_stats + def batch_norm_param(self, use_global_stats=0, moving_average_fraction=None, eps=None): + bn_param = pb.BatchNormParameter() + bn_param.use_global_stats = use_global_stats if moving_average_fraction: - bn_param.moving_average_fraction=moving_average_fraction + bn_param.moving_average_fraction = moving_average_fraction if eps: bn_param.eps = eps self.param.batch_norm_param.CopyFrom(bn_param) - # layer - # { - # name: "upsample_layer" - # type: "Upsample" - # bottom: "some_input_feature_map" - # bottom: "some_input_pool_index" - # top: "some_output" - # upsample_param { - # upsample_h: 224 - # upsample_w: 224 - # } - # } - def upsample_param(self,size=None, scale_factor=None): - upsample_param=pb.UpsampleParameter() + def upsample_param(self, size=None, scale_factor=None): + upsample_param = pb.UpsampleParameter() if scale_factor: - if isinstance(scale_factor,int): + if isinstance(scale_factor, int): upsample_param.scale = scale_factor else: upsample_param.scale_h = scale_factor[0] upsample_param.scale_w = scale_factor[1] if size: - if isinstance(size,int): + if isinstance(size, int): upsample_param.upsample_h = size else: upsample_param.upsample_h = size[0] upsample_param.upsample_w = size[1] - #upsample_param.upsample_h = size[0] * scale_factor - #upsample_param.upsample_w = size[1] * scale_factor + # upsample_param.upsample_h = size[0] * scale_factor + # upsample_param.upsample_w = size[1] * scale_factor self.param.upsample_param.CopyFrom(upsample_param) - def interp_param(self,size=None, scale_factor=None): - interp_param=pb.InterpParameter() + + def interp_param(self, size=None, scale_factor=None): + interp_param = pb.InterpParameter() if scale_factor: - if isinstance(scale_factor,int): + if isinstance(scale_factor, int): interp_param.zoom_factor = scale_factor if size: @@ -138,7 +130,7 @@ def interp_param(self,size=None, scale_factor=None): interp_param.width = size[1] self.param.interp_param.CopyFrom(interp_param) - def add_data(self,*args): + def add_data(self, *args): """Args are data numpy array """ del self.param.blobs[:] @@ -148,11 +140,12 @@ def add_data(self,*args): new_blob.shape.dim.append(dim) new_blob.data.extend(data.flatten().astype(float)) - def set_params_by_dict(self,dic): + def set_params_by_dict(self, dic): pass - def copy_from(self,layer_param): + def copy_from(self, layer_param): pass -def set_enum(param,key,value): - setattr(param,key,param.Value(value)) + +def set_enum(param, key, value): + setattr(param, key, param.Value(value)) diff --git a/tools/deploy/Caffe/net.py b/tools/deploy/Caffe/net.py index d2291f576..3a219d6a4 100644 --- a/tools/deploy/Caffe/net.py +++ b/tools/deploy/Caffe/net.py @@ -1 +1,2 @@ -raise ImportError,'the nn_tools.Caffe.net is no longer used, please use nn_tools.Caffe.caffe_net' \ No newline at end of file +raise ImportError("the nn_tools.Caffe.net is no longer used, please use nn_tools.Caffe.caffe_net") +