Have a personal or library account? Click to login
Deep Learning with 3D ResNets for Comprehensive Dual-Lane Speed Climbing Video Analysis Cover

Deep Learning with 3D ResNets for Comprehensive Dual-Lane Speed Climbing Video Analysis

By: Y. Xie and  V. Mariano  
Open Access
|Mar 2025

Figures & Tables

Figure 1.

The International Federation of Sport Climbing (IFSC) World Cup Climbing Men’s Speed Climbing Competition site, Chamonix, France, 2018. Photograph by Jan Kriz, via Wikimedia Commons. (https://en.wikipedia.org/wiki/Speed_climbing_wall).
The International Federation of Sport Climbing (IFSC) World Cup Climbing Men’s Speed Climbing Competition site, Chamonix, France, 2018. Photograph by Jan Kriz, via Wikimedia Commons. (https://en.wikipedia.org/wiki/Speed_climbing_wall).

Figure 2.

Diagram of the official speed climb wall including numbered hand holds. Adapted from (Walltopia 2020). *Final button or ‘hold 21’ (Lau, 2021)
Diagram of the official speed climb wall including numbered hand holds. Adapted from (Walltopia 2020). *Final button or ‘hold 21’ (Lau, 2021)

Figure 3.

Images of the three training situations for speed climbing, (1)Dual-Lane double climbing, both lanes are occupied, (2)Dual-Lane single climbing on the left lane, the left lane is occupied and the right lane is empty, (3)Dual-Lane single climbing on the right lane, the right lane is occupied and the left lane is empty.
Images of the three training situations for speed climbing, (1)Dual-Lane double climbing, both lanes are occupied, (2)Dual-Lane single climbing on the left lane, the left lane is occupied and the right lane is empty, (3)Dual-Lane single climbing on the right lane, the right lane is occupied and the left lane is empty.

Figure 4.

Images of the Three States in Speed Climbing: Flash, Slip, and Fall, (1)Flash, the climber has tapped the timer to stop and the display light is green, (2)Slip, the climber’s left foot steps out of the air and slips (3)Fall, he climber’s feet are dangling in the air, falling downward slowly, and the timer continues to keep time and the display light is red.
Images of the Three States in Speed Climbing: Flash, Slip, and Fall, (1)Flash, the climber has tapped the timer to stop and the display light is green, (2)Slip, the climber’s left foot steps out of the air and slips (3)Fall, he climber’s feet are dangling in the air, falling downward slowly, and the timer continues to keep time and the display light is red.

Figure 5.

Simplified diagram of the 3D ResNet model architecture designed for this study.
Simplified diagram of the 3D ResNet model architecture designed for this study.

Figure 6.

Accuracy Curve of 3D ResNet Model. The horizontal “Epoch” indicates the number of training rounds, starting from 0 and incrementing, demonstrating the model’s training iteration. The vertical “Accuracy” represents the model’s accuracy in classifying 15 climbing results on the Training and Testing Accuracy sets, ranging from 0 to 1.
Accuracy Curve of 3D ResNet Model. The horizontal “Epoch” indicates the number of training rounds, starting from 0 and incrementing, demonstrating the model’s training iteration. The vertical “Accuracy” represents the model’s accuracy in classifying 15 climbing results on the Training and Testing Accuracy sets, ranging from 0 to 1.

Figure 7.

Loss Curve of 3D ResNet Model. The horizontal coordinate is also “Epoch” and the vertical coordinate “Loss” indicates the loss value of the model on the training set and the Testing set, the magnitude of which reflects the model’s predicted the degree of difference between the results and the true labels.
Loss Curve of 3D ResNet Model. The horizontal coordinate is also “Epoch” and the vertical coordinate “Loss” indicates the loss value of the model on the training set and the Testing set, the magnitude of which reflects the model’s predicted the degree of difference between the results and the true labels.

Figure 8.

Confusion Matrix for 3D ResNet Model Performance on Speed Climbing Video Analysis. Rows: Represent the actual classes. Columns: Represent the predicted classes. Diagonal Values: Indicate correct classifications, with higher values reflecting better performance. Off-Diagonal Values: Represent misclassifications, identifying areas where the model struggled.
Confusion Matrix for 3D ResNet Model Performance on Speed Climbing Video Analysis. Rows: Represent the actual classes. Columns: Represent the predicted classes. Diagonal Values: Indicate correct classifications, with higher values reflecting better performance. Off-Diagonal Values: Represent misclassifications, identifying areas where the model struggled.

Comparison table for classification, labelling and coding of video status for dual lane speed climbing_

Dual-lane climb stateAnnotationEncodeVideos Quantities
left flash, right flash1-10237
left flash, right slip1-2186
left flash, right fall1-3257
left flash, right empty1-4372
left slip, right flash2-1493
left slip, right slip2-2545
left slip, right fall2-3620
left slip, right empty2-4731
left fall, right flash3-1844
left fall, right slip3-2910
left fall, right fall3-31029
left fall, right empty3-41115
left empty, right flash4-11290
left empty, right slip4-21320
left empty, right fall4-31423

Table of 3D ResNet model classification report

ClassPrecisionRecallF1-ScoreSupport
0 (L-Flash; R-Flash)0.910.950.93594
1 (L-Flash; R-Slip)0.850.910.88190
2 (L-Flash; R-Fall)0.950.820.88131
3 (L-Flash; R-Empty)0.950.960.95164
4 (L-Slip; R-Flash)0.940.820.88244
5 (L-Slip; R-Slip)0.930.920.93125
6 (L-Slip; R-Fall)0.890.980.9357
7 (L-Slip; R-Empty)0.950.940.9583
8 (L-Fall; R-Flash)0.930.930.93124
9 (L-Fall; R-Slip)0.950.950.9520
10 (L-Fall; R-Fall)0.840.910.8858
11 (L-Fall; R-Empty)1.000.790.8824
12 (L-Empty; R-Flash)0.990.980.99250
13 (L-Empty; R-Slip)0.990.990.9972
14 (L-Empty; R-Fall)0.981.000.9952

Performance Comparison of 3D ResNet, 2D CNN and C3D in Terms of Accuracy and Loss_

ModelAccuracyLoss
3D ResNet92.78%0.57
2D CNN25.62%2.42
C3D27.15%2.51
Language: English
Page range: 17 - 34
Published on: Mar 2, 2025
Published by: International Association of Computer Science in Sport
In partnership with: Paradigm Publishing Services
Publication frequency: 2 issues per year

© 2025 Y. Xie, V. Mariano, published by International Association of Computer Science in Sport
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.