Focal Loss for Dense Object Detection

By Tsung-Yi Lin, Priya Goyal, Ross B. Girshick, Kaiming He, Piotr Doll'ar. 2017

32,959 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Multi-Granularity Hierarchical Attention Fusion Networks for Reading Comprehension and Question Answering (2018, methodology): “During ﬁne-tuning, Focal Loss (Lin et al., 2017) and Reinforce Loss which take F1 score as reward are incorporated with Cross Entropy Loss.”
End-to-End Object Detection with Transformers (2020, methodology): “The final resolution of the masks has stride 4 and each mask is supervised independently using the DICE/F-1 loss [28] and Focal loss [23].”
Big Transfer (BiT): General Visual Representation Learning (2019, methodology): “Model Upstream data AP RetinaNet [33] ILSVRC-2012 40.”
Stand-Alone Self-Attention in Vision Models (2019, background): “The impressive performance of fully attentional models veriﬁes that stand-alone attention is a viable primitive for vision models.”
Attention Augmented Convolutional Networks (2019, methodology): “ResNet-50 and its larger counterparts use a bottleneck block comprising of 1x1, 3x3, 1x1 convolutions where the last pointwise convolution expands the number of ﬁlters and the ﬁrst one contracts the number of ﬁlters.”
Group Normalization (2018, background): “These computer vision tasks in general benefit from higher-resolution input, so the batch size tends to be small in common practice (1 or 2 images/GPU [12, 47, 18, 36]).”
Relation Networks for Object Detection (2017, neutral): “3Another object detection paradigm is based on dense sliding windows [35, 37, 33].”
Randaugment: Practical automated data augmentation with a reduced search space (2019, methodology): “Brieﬂy, we employed RetinaNet [26] with ResNet-101 and ResNet-200 as a backbone [17].”
Cascade R-CNN: High Quality Object Detection and Instance Segmentation (2019, methodology): “G-RMI [28], the recent Deformable R-FCN [8], RetinaNet [36], Mask R-CNN [24], RelationNet [27], DetNet [34], CornerNet [32], etc.”
Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression (2019, background): “The dominant trend for improving performance of applications utilizing deep neural networks is to propose either a better architecture backbone [15, 13] or a better strategy to extract reliable local features [6].”
Make-A-Scene: Scene-Based Text-to-Image Generation with Human Priors (2022, methodology): “A straightforward solution would be to employ a loss more suitable for class imbalance, such as focal loss [36].”
MegDet: A Large Mini-Batch Object Detector (2017, methodology): “The improvements are mainly due to better backbone network [16], new detection framework [31], novel loss design [24], improved pooling method [5, 14], and so on [19].”
On the Evaluation of Vision-and-Language Navigation Instructions (2021, methodology): “For the classification loss Lcls we consider both binary cross entropy loss LCE, and focal loss (Lin et al., 2017) given by LFL = (1 − pi,j)γLCE where we set γ = 2.”
Segment Anything (2023, positive): “We supervise mask prediction with the linear combination of focal loss [65] and dice loss [73] used in [14].”
Pix2seq: A Language Modeling Framework for Object Detection (2021, methodology): “For example, the loss function is often based on bounding box regression (Szegedy et al., 2013; Lin et al., 2017b), intersection over union (Rezatoﬁghi et al., 2019), and set-based matching (Erhan et al., 2014; Liu et al., 2016; Redmon et al., 2016; Stewart et al., 2016; Carion et al., 2020).”
Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection (2023, methodology): “Specifically, we dot product each query with text features to predict logits for each text token and then compute focal loss [28] for each logit.”
Exploring Plain Vision Transformer Backbones for Object Detection (2022, methodology): “Figure 5 is the RetinaNet [38] counterpart of Figure 3, showing the trade-off between accuracy and model size.”
Instance Segmentation for Autonomous Log Grasping in Forestry Operations (2022, background): “One example is YOLACT++ [11], which is extending RetinaNet’s [12] dense grid prediction by adding a mask branch that combines prototype masks extracted from the feature maps.”