Amulet: Aggregating Multi-level Convolutional Features for Salient Object Detection
Citations
1,553 citations
962 citations
Cites background or methods from "Amulet: Aggregating Multi-level Con..."
...Zhang et al. (UCF) [75] developed a reformulated dropout and a hybrid upsampling module to reduce the checkboard artifacts of deconvolution operators as well as aggregating multi-level convolutional features in (Amulet) [74] for saliency detection....
[...]
...(UCF) [75] developed a reformulated dropout and a hybrid upsampling module to reduce the checkboard artifacts of deconvolution operators as well as aggregating multi-level convolutional features in (Amulet) [74] for saliency detection....
[...]
...Amulet: Aggregating multi-level convolutional features for salient object detection....
[...]
...We compare our method with 15 state-of-the-art models, PiCANetR [39], BMPM [72], R(3)Net [6], PAGRN [76], RADF [19], DGRL [65], RAS [4], C2S [36], LFR [73], DSS [17], NLDF [41], SRM [64], Amulet [74], UCF [75], MDF [35]....
[...]
...We compare our method with 15 state-of-the-art models, PiCANetR [39], BMPM [72], R3Net [6], PAGRN [76], RADF [19], DGRL [65], RAS [4], C2S [36], LFR [73], DSS [17], NLDF [41], SRM [64], Amulet [74], UCF [75], MDF [35]....
[...]
803 citations
759 citations
Cites background or methods from "Amulet: Aggregating Multi-level Con..."
...When the edge branch is not incorporated, training only takes less than 6 hours on a training set of 5,000 images, which is quite faster than most of the previous methods [24, 43, 28, 44, 45, 9]....
[...]
...As pointed out in many previous approaches [9, 28, 44], because of the pyramid-like structural characteristics of CNNs, shallower stages usually have larger spatial sizes and keep rich, detailed low-level information while deeper stages contain more high-level semantic knowledge and are better at locating the exact places of salient objects....
[...]
...[44] both advanced the Ushape structures and utilized multiple levels of context in-...
[...]
...Image GT Ours PiCANet[24] DGRL[38] PAGR[46] SRM[37] Amulet[44] DSS[9] MSR[17] DCL[19]...
[...]
...Ours PiCANet [24] DGRL [38] SRM [37] Amulet [44] Size 400× 300 224× 224 384× 384 353× 353 256× 256...
[...]
758 citations
Cites background or methods from "Amulet: Aggregating Multi-level Con..."
...We implement the improved models in their respectively default deep learning library (tensorflow [1] for BMPM and NLDF, caffe [12] for Amulet)....
[...]
...In this paper, we apply the proposed framework in three deep aggregation models (BMPM, Amulet, NLDF)....
[...]
...And we re-train NLDF, DSS, BMPM on this dataset....
[...]
...For BMPM and NLDF, we train the improved models (denoted as BMPM-CPD and NLDFCPD) by using default settings, and it only needs to change the learning rate from the original 10−6 to 10−5....
[...]
...For A- mulet, we train the improved model (denoted as AmuletCPD) by using the completely same settings as the original model....
[...]
References
123,388 citations
55,235 citations
"Amulet: Aggregating Multi-level Con..." refers methods in this paper
...All compared methods are based on the same VGG-16 model pre-trained on the ImageNet classification task [37]....
[...]
...We build our architecture on the VGG-16 model from [37], which is well known for its elegance and simplicity, and at the same time yields nearly state-of-the-art results in image classification and good generalization properties....
[...]
...The parameters of multilevel feature extraction layers are initialized from the VGG16 model [37]....
[...]
...Given an input image (256×256×3), multi-level features are first generated by the feature extraction network (VGG-16 [37])....
[...]
49,914 citations
49,590 citations
28,225 citations