1/119
Looks like no tags are added yet.
accommodate /əˈkɑməˌdeɪt/
【動詞】対応する、収める | 例文:Furthermore, RT-DETR supports flexible speed tuning to accommodate various real-time scenarios without retraining, thanks to the multi-layer decoder architecture of DETR.(Introduction)
achieves /əˈtʃiːvz/
【動詞】達成する(3人称単数現在)(基本形:achieve) | 例文:Our RT-DETR-R50 / R101 achieves 53.1% / 54.3% AP on COCO and 108 / 74 FPS on T4 GPU, outperforming previously advanced YOLOs in both speed and accuracy.(Abstract)
adjusting /əˈdʒʌstɪŋ/
【動詞】調整する(〜ing形)(基本形:adjust) | 例文:In addition, RT-DETR supports flexible speed tuning by adjusting the number of decoder layers to adapt to various scenarios without retraining.(Abstract)
adopt /əˈdɑpt/
【動詞】採用する | 例文:However, we observe that the current query selection directly adopt classification scores for selection, ignoring the fact that the detector are required to simultaneously model the category and location of objects, both of which determine the quality of the features.(Introduction)
analyze /ˈænəˌlaɪz/
【動詞】分析する、解析する | 例文:We quantitatively analyze the impact of NMS on the speed and accuracy of YOLO detectors, and establish an end-to-end speed benchmark to test the end-to-end inference speed of real-time detectors.(Introduction)
compare /kəmˈpɛr/
【動詞】比較する | 例文:We compare the scaled RT-DETRs with the lighter (S and M ) YOLO detectors in Appendix, which outperform all S and M models in both speed and accuracy.(Method)
concatenate /kənˈkætəˌneɪt/
【動詞】連結する(チャンネル方向に結合する) | 例文:The multi-scale features share the encoder for intra-scale feature interaction and then concatenate as output.(Method)
conclude /kənˈklud/
【動詞】結論づける、結論を出す | 例文:From the results, we can conclude that the execution time of the EfficientNMS kernel increases as the confidence threshold decreases or the IoU threshold increases.(End-to-end Speed of Detectors)
conduct /kənˈdʌkt/
【動詞】実施する、行う | 例文:To achieve the above goal, we rethink DETRs and conduct detailed analysis of key components to reduce unnecessary computational redundancy and further improve accuracy.(Introduction)
decouples /ˌdiˈkʌpəlz/
【動詞】分離する、切り離す(基本形:decouple) | 例文:Variant D decouples intra-scale interaction and cross-scale fusion by utilizing the single-scale Transformer encoder for the former and a PANet-style [22] structure for the latter.(Method)
enhances /ɛnˈhænsɪz/
【動詞】向上させる、強化する(基本形:enhance) | 例文:Lite DETR [18] enhances the efficiency of encoder by reducing the update frequency of low-level features in an interleaved way.(Related Work)
establish /ɪˈstæblɪʃ/
【動詞】確立する、示す | 例文:We quantitatively analyze the impact of NMS on the speed and accuracy of YOLO detectors, and establish an end-to-end speed benchmark to test the end-to-end inference speed of real-time detectors.(Introduction)
evaluate /ɪˈvæljəˌweɪt/
【動詞】評価する、検証する | 例文:Furthermore, we use YOLOv8 to evaluate the accuracy on the COCO val2017 and test the execution time of the NMS operation under different hyperparameters.(End-to-end Speed of Detectors)
exceeds /ɪkˈsiːdz/
【動詞】超える、上回る(基本形:exceed) | 例文:Specifically, the boxes with scores below the confidence threshold are directly filtered out, and whenever the IoU of any two boxes exceeds the IoU threshold, the box with the lower score will be discarded.(End-to-end Speed of Detectors)
flatten /ˈflætən/
【動詞】平坦化する(1次元に展開する) | 例文:We formulate the calculation of the hybrid encoder as: Q = K = V = Flatten(S5), F5 = Reshape(AIFI(Q, K, V)), O = CCFF({S3, S4, F5}), where Reshape represents restoring the shape of the flattened feature to the same shape as S5.(Method)
generate /ˈdʒɛnəˌreɪt/
【動詞】生成する、生み出す | 例文:Finally, the decoder with auxiliary prediction heads iteratively optimizes object queries to generate categories and boxes.(Method)
improves /ɪmˈpruːvz/
【動詞】改善する(3人称単数現在)(基本形:improve) | 例文:The introduction of multi-scale features accelerates training convergence and improves performance [45].(Method)
increases /ɪnˈkriːsɪz/
【動詞】増える、増加する(基本形:increase) | 例文:As the confidence threshold increases, more prediction boxes are filtered out, and the number of remaining boxes that need to calculate IoU decreases, thus reducing the execution time of NMS.(End-to-end Speed of Detectors)
introduces /ˌɪntrəˈdusɪz/
【動詞】導入する、提示する(基本形:introduce) | 例文:However, these detectors typically require Non-Maximum Suppression (NMS) for post-processing, which not only slows down the inference speed but also introduces hyperparameters that cause instability in both the speed and accuracy.(Introduction)
minimize /ˈmɪnɪmaɪz/
【動詞】最小化する | 例文:To minimize the uncertainty of the queries, we integrate the uncertainty into the loss function for the gradient-based optimization in Eq. (3).(Method)
observe /əbˈzɝv/
【動詞】観察する、認める | 例文:However, we observe that the speed and accuracy of YOLOs are negatively affected by the NMS.(Abstract)
overcome /ˌoʊvərˈkʌm/
【動詞】克服する、乗り越える | 例文:To overcome this bottleneck, we first analyze the computational redundancy present in the multi-scale Transformer encoder.(Method)
propose /prəˈpoʊz/
【動詞】提案する | 例文:In this paper, we propose the Real-Time DEtection TRansformer (RT-DETR), the first real-time end-to-end object detector to our best knowledge that addresses the above dilemma.(Abstract)
reduces /rɪˈduːsɪz/
【動詞】減らす、削減する(基本形:reduce) | 例文:However, although the deformable attention reduces the computational cost, the sharply increased sequence length still causes the encoder to become the computational bottleneck.(Method)
refers /rɪˈfɜːrz/
【動詞】指す、言及する(「〜を指す」の現在形)(基本形:refer) | 例文:Note that the NMS operation we adopt refers to the TensorRT efficientNMSPlugin, which involves multiple kernels, including EfficientNMSFilter, RadixSort, EfficientNMS, etc., and we only report the execution time of the EfficientNMS kernel.(End-to-end Speed of Detectors)
replace /rɪˈpleɪs/
【動詞】置き換える、置換する | 例文:And for the latter, previous works [42, 44, 45] show that the hard-to-optimize object queries hinder the performance of DETRs and propose the query selection schemes to replace the vanilla learnable embeddings with encoder features.(Introduction)
represents /ˌrɛprɪˈzɛnts/
【動詞】表す、示す(基本形:represent) | 例文:SSE represents the single-scale Transformer encoder, MSE represents the multi-scale Transformer encoder, and CSF represents cross-scale fusion.(End-to-end Speed of Detectors)
require /rɪˈkwaɪər/
【動詞】必要とする、要求する | 例文:However, these detectors typically require Non-Maximum Suppression (NMS) for post-processing, which not only slows down the inference speed but also introduces hyperparameters that cause instability in both the speed and accuracy.(Introduction)
reshape /riˈʃeɪp/
【動詞】形状を変える、リシェイプする | 例文:O = CCFF({S3 , S4 , F5 }), where Reshape represents restoring the shape of the flattened feature to the same shape as S5 .(Method)
rethink /riˈθɪŋk/
【動詞】再考する、見直す | 例文:To achieve the above goal, we rethink DETRs and conduct detailed analysis of key components to reduce unnecessary computational redundancy and further improve accuracy.(Introduction)
select /səˈlɛkt/
【動詞】選択する | 例文:Moreover, considering that different scenarios place different emphasis on recall and accuracy, it is necessary to carefully select the appropriate NMS thresholds, which hinders the development of real-time detectors.(Introduction)
shows /ʃoʊz/
【動詞】示す(3人称単数現在)(基本形:show) | 例文:This shows that uncertainty-minimal query selection produces more high-quality encoder features.(Experiments)
suffers /ˈsʌfərz/
【動詞】(悪影響を)被る、悩まされる(基本形:suffer) | 例文:Despite its obvious advantages, DETR suffers from several problems: slow training convergence, high computational cost, and hard-to-optimize queries.(Related Work)
supports /səˈpɔːrts/
【動詞】対応する、サポートする(基本形:support) | 例文:In addition, RT-DETR supports flexible speed tuning by adjusting the number of decoder layers to adapt to various scenarios without retraining.(Abstract)
utilize /ˈjuːtəlaɪz/
【動詞】活用する、利用する | 例文:We utilize the benchmark to test the end-to-end speed of anchor-based detectors YOLOv5 [11] and YOLOv7 [38], as well as anchor-free detectors PP-YOLOE [40], YOLOv6 [16] and YOLOv8 [12] on T4 GPU with TensorRT FP16.(End-to-end Speed of Detectors)
visualize /ˈvɪʒuəˌlaɪz/
【動詞】可視化する | 例文:We also visualize the predictions of YOLOv8 with different NMS thresholds in Appendix.(End-to-end Speed of Detectors)
ablation /əˈbleɪʃən/
【名詞】アブレーション(要素実験) | 例文:We conduct an ablation study on uncertainty-minimal query selection, and the results are reported on RT-DETR-R50 with 1× configuration, Table 4.(Experiments)
analysis /əˈnæləsɪs/
【名詞】分析、解析 | 例文:To achieve the above goal, we rethink DETRs and conduct detailed analysis of key components to reduce unnecessary computational redundancy and further improve accuracy.(Introduction)
anchor /ˈæŋkər/
【名詞】アンカー(基準枠) | 例文:Particularly, DETR eliminates the hand-crafted anchor and NMS components.(Related Work)
approach /əˈproʊtʃ/
【名詞】手法、アプローチ/接近する | 例文:RT-DETR, along with its model scaling strategy, broadens the technical approach to real-time object detection, offering new possibilities beyond YOLO for diverse real-time scenarios.(Conclusion)
attention /əˈtɛnʃən/
【名詞】注意機構、アテンション | 例文:Carion et al. [4] first propose the end-to-end detector based on Transformer called DETR, which has attracted extensive attention due to its distinctive features.(Related Work)
backbone /ˈbækboʊn/
【名詞】バックボーン(特徴抽出部) | 例文:RT-DETR consists of a backbone, an efficient hybrid encoder, and a Transformer decoder with auxiliary prediction heads.(Method)
baseline /ˈbeɪslaɪn/
【名詞】基準、ベースライン(比較対象の標準手法・値) | 例文:Compared to baseline A, variant B improves accuracy by 1.9% AP and increases the latency by 54%.(Experiments)
benchmark /ˈbɛntʃmɑːrk/
【名詞】ベンチマーク(比較評価の基準) | 例文:We quantitatively analyze the impact of NMS on the speed and accuracy of YOLO detectors, and establish an end-to-end speed benchmark to test the end-to-end inference speed of real-time detectors.(Introduction)
breakthrough /ˈbreɪkθruː/
【名詞】画期的な進展、ブレークスルー | 例文:We view query initialization as a breakthrough to further improve performance.(Introduction)
category /ˈkætəˌɡɔri/
【名詞】カテゴリ、分類 | 例文:This process is performed iteratively until all boxes of every category have been processed.(End-to-end Speed of Detectors)
comparison /kəmˈpærɪsən/
【名詞】比較 | 例文:To enable a fair comparison of the end-to-end speed of various real-time detectors, we establish an end-to-end speed benchmark.(End-to-end Speed of Detectors)
confidence /ˈkɑːnfɪdəns/
【名詞】信頼度、確信度 | 例文:The confidence score represents the likelihood that the feature includes foreground objects.(Method)
configuration /kənˌfɪɡjəˈreɪʃən/
【名詞】構成、設定 | 例文:We evaluate the indicators of the variants designed in Sec. 4.2, including AP (trained with 1× configuration), the number of parameters, and the latency, Table 3.(Experiments)
cost /kɔːst/
【名詞】コスト、計算量・処理負担 | 例文:Nevertheless, the high computational cost limits their practicality and hinders them from fully exploiting the advantage of excluding NMS.(Abstract)
difficulty /ˈdɪfəkəlti/
【名詞】難しさ、困難 | 例文:Conditional DETR [27] and Anchor DETR [39] decrease the optimization difficulty of the queries.(Related Work)
effectiveness /ɪˈfɛktɪvnəs/
【名詞】有効性、効果 | 例文:To analyze the effectiveness of the uncertainty-minimal query selection, we visualize the classification scores and IoU scores of the selected features on COCO val2017, Figure 6.(Method)
embedding /ɪmˈbɛdɪŋ/
【名詞】埋め込み(特徴ベクトル) | 例文:Specifically, for the hybrid encoder, we control the width by adjusting the embedding dimension and the number of channels, and the depth by adjusting the number of Transformer layers and RepBlocks.(Experiments)
foreground /ˈfɔrˌɡraʊnd/
【名詞】前景(画像中の対象物体がある領域) | 例文:The confidence score represents the likelihood that the feature includes foreground objects.(Method)
fusion /ˈfjuːʒən/
【名詞】融合、統合 | 例文:CCFF is optimized based on the cross-scale fusion module, which inserts several fusion blocks consisting of convolutional layers into the fusion path.(Method)
hyperparameters /ˌhaɪpərpəˈræmətərz/
【名詞】ハイパーパラメータ(基本形:hyperparameter) | 例文:However, these detectors typically require Non-Maximum Suppression (NMS) for post-processing, which not only slows down the inference speed but also introduces hyperparameters that cause instability in both the speed and accuracy.(Introduction)
improvement /ɪmˈpruvmənt/
【名詞】改善、向上 | 例文:Variant C delivers a 0.7% AP improvement over B and increases the latency by 20%.(Experiments)
inference /ˈɪnfərəns/
【名詞】推論 | 例文:However, these detectors typically require Non-Maximum Suppression (NMS) for post-processing, which not only slows down the inference speed but also introduces hyperparameters that cause instability in both the speed and accuracy.(Introduction)
latency /ˈleɪtənsi/
【名詞】遅延、レイテンシ | 例文:Variant C delivers a 0.7% AP improvement over B and increases the latency by 20%.(Experiments)
limitation /ˌlɪməˈteɪʃən/
【名詞】制限、限界、制約 | 例文:Although the proposed RT-DETR outperforms the state-of-the-art real-time detectors and end-to-end detectors with similar size in both speed and accuracy, it shares the same limitation as the other DETRs, i.e., the performance on small objects is still inferior than the strong real-time detectors.(Limitation and Discussion)