From 45a73c23d39f6e0b598371cd0356c4a2f50d1f1f Mon Sep 17 00:00:00 2001 From: Daniel Bolya Date: Mon, 6 Aug 2018 15:41:40 -0700 Subject: [PATCH] Added the ability to select the config to use in the eval script. --- data/config.py | 17 ++++++----------- eval.py | 8 +++++++- yolact.py | 7 +++++-- 3 files changed, 18 insertions(+), 14 deletions(-) diff --git a/data/config.py b/data/config.py index 81e61db1c..2480278c0 100644 --- a/data/config.py +++ b/data/config.py @@ -66,13 +66,9 @@ def replace(self, new_config_dict): 'type': ResNetBackbone, 'args': ([3, 4, 23, 3],), - 'selected_layers': list(range(2, 7)), - 'pred_scales': [[1, 2], [2], [2], [2], [2]], - 'pred_aspect_ratios': [[[1], [1.05, 0.62]], - [[1.29, 0.79, 0.47, 2.33, 0.27]], - [[1.19, 0.72, 0.43, 2.13, 0.25]], - [[1.34, 0.84, 0.52, 2.38, 0.30]], - [[1.40, 0.95, 0.64, 2.16]]], + 'selected_layers': list(range(2, 8)), + 'pred_scales': [[1]]*6, + 'pred_aspect_ratios': [ [[0.66685089, 1.7073535, 0.87508774, 1.16524493, 0.49059086]] ] * 6, }) resnet50_backbone = resnet101_backbone.copy({ @@ -143,11 +139,10 @@ def replace(self, new_config_dict): 'name': 'yolact_resnet101', 'backbone': resnet101_backbone, - 'min_size': 400, - 'max_size': 600, + 'max_size': 550, 'train_masks': True, - 'preserve_aspect_ratio': True, + 'preserve_aspect_ratio': False, 'use_prediction_module': True, 'use_yolo_regressors': True, }) @@ -161,7 +156,7 @@ def replace(self, new_config_dict): 'pred_aspect_ratios': [ [[1], [1, sqrt(2), 1/sqrt(2), sqrt(3), 1/sqrt(3)][:n]] for n in [3, 5, 5, 5, 3, 3] ], }), - 'max_size': 600, + 'max_size': 550, 'train_masks': True, 'preserve_aspect_ratio': False, diff --git a/eval.py b/eval.py index 639127463..aa71ba2c8 100644 --- a/eval.py +++ b/eval.py @@ -7,7 +7,7 @@ from utils.functions import sanitize_coordinates, SavePath import pycocotools -from data import cfg +from data import cfg, set_cfg import numpy as np import torch @@ -73,11 +73,16 @@ def str2bool(v): help='The output file for coco mask results if --coco_results is set.') parser.add_argument('--max_num_detections', default=100, type=int, help='The maximum number of detections to consider for each image for mAP scoring. COCO uses 100.') +parser.add_argument('--config', default=None, + help='The config object to use.') parser.set_defaults(display=False, resume=False, output_coco_json=False, shuffle=False) args = parser.parse_args() +if args.config is not None: + set_cfg(args.config) + iou_thresholds = [x / 100 for x in range(50, 100, 5)] coco_cats = [] # Call prep_coco_cats to fill this @@ -577,6 +582,7 @@ def evaluate(net, dataset): if it > 1: print('Avg FPS: %.4f' % (1 / frame_times.get_avg())) plt.imshow(np.clip(img_numpy, 0, 1)) + plt.title(str(dataset.ids[image_idx])) plt.show() else: if it > 1: fps = 1 / frame_times.get_avg() diff --git a/yolact.py b/yolact.py index 629ad93ab..14c9af753 100644 --- a/yolact.py +++ b/yolact.py @@ -118,7 +118,7 @@ def make_priors(self, conv_h, conv_w): # +0.5 because priors are in center-size notation x = (i + 0.5) / conv_w y = (j + 0.5) / conv_h - + for scale, ars in zip(self.scales, self.aspect_ratios): for ar in ars: w = scale * ar / conv_w @@ -229,8 +229,11 @@ def forward(self, x): # torch.set_default_tensor_type('torch.cuda.FloatTensor') x = torch.zeros((1, 3, cfg.max_size, cfg.max_size)) - y = net(x) + + for p in net.prediction_layers: + print(p.last_conv_size) + print() for a in y: print(a.size(), torch.sum(a))