This submit is co-written with Tom Koerick from SkillShow.
The youth sports activities market was valued at $37.5 billion globally in 2022 and is projected to develop by 9.2% annually via 2030. Roughly 60 million younger athletes take part on this market worldwide. SkillShow, a pacesetter in youth sports activities video manufacturing, movies over 300 occasions yearly within the youth sports activities business, creating content material for over 20,000 younger athletes yearly. This submit describes how SkillShow used Amazon Transcribe and different Amazon Internet Companies (AWS) machine studying (ML) providers to automate their video processing workflow, lowering enhancing time and prices whereas scaling their operations.
Problem
In response to the surge in youth sports activities video manufacturing, guide video enhancing processes have gotten more and more unsustainable. Since 2001, SkillShow has been on the forefront of sports activities video manufacturing, offering complete video providers for people, groups, and occasion organizers. They focus on filming, enhancing, and distributing content material that helps athletes showcase their abilities to recruiters, construct their private model on social media, and help their growth coaching. As a trusted companion to main sports activities organizations together with the Excellent Sport, 3Step Sports activities, USA Baseball, MLB Community, Underneath Armour, Elite11 soccer combines and extra, SkillShow has filmed a whole lot of 1000’s of athletes and 1000’s of regional and nationwide occasions throughout completely different sports activities and age teams.
Regardless of their market management, SkillShow confronted vital operational challenges. With solely seven full-time staff managing their increasing operation, they needed to outsource to over 1,100 contractors yearly. This reliance on outsourced enhancing not solely elevated operational prices but additionally resulted in a prolonged 3-week turnaround time per occasion, making it troublesome to maintain tempo with the rising demand for youth sports activities content material.
Managing roughly 230 TB of video knowledge per 12 months created vital operational challenges. This huge quantity of information meant prolonged add and obtain occasions for editors, costly storage prices, and complicated knowledge administration necessities. Every occasion’s uncooked footage wanted to be securely saved, backed up, and made accessible to a number of editors, straining each technical sources and IT infrastructure. These challenges led to SkillShow halting new occasions mid-2023, limiting their progress potential in a quickly increasing market. The necessity for an environment friendly, scalable answer grew to become important to sustaining SkillShow’s place and assembly the rising demand for youth sports activities content material, significantly within the post-COVID period the place recruiting movies have turn into important for leagues and athletes alike.
Resolution overview
To handle these challenges, SkillShow partnered with AWS to develop an automatic video processing pipeline. The staff initially explored a number of approaches to automate participant identification.
Facial recognition proved difficult because of various video high quality, inconsistent lighting situations, and frequent participant motion throughout video games. Moreover, gamers usually wore tools equivalent to helmets or protecting gear that obscured their faces, making dependable identification troublesome.
Textual content-based detection of jersey numbers and colours appeared promising at first, however offered its personal set of challenges. Jersey numbers had been continuously obscured by participant motion, climate situations may have an effect on visibility, and ranging digital camera angles made constant detection unreliable.
Finally, the staff settled on an audio logging and automatic clip technology answer, which proved superior for a number of causes:
- Extra dependable participant identification, as a result of announcers constantly name out participant numbers and staff colours
- Higher efficiency in various environmental situations, as a result of audio high quality stays comparatively constant even in difficult climate or lighting
- Decreased processing complexity and computational necessities in comparison with video-based evaluation
- Cheaper because of decrease computational calls for and better accuracy charges
- Capability to seize extra context from announcer commentary, equivalent to play descriptions and recreation conditions
This answer makes use of a number of key AWS providers:
- Amazon Easy Storage Service (Amazon S3):
- Used for storing the enter and output video information
- Offers scalable and sturdy storage to deal with SkillShow’s giant video knowledge quantity of 230 TB per 12 months
- Permits for easy entry and integration with different AWS providers within the processing pipeline
- AWS Lambda:
- Serverless compute service used to energy the automated processing workflows
- Triggers the varied capabilities that orchestrate the video processing, equivalent to transcription and clip technology
- Allows event-driven, scalable, and cost-effective processing with out the necessity to handle underlying infrastructure
- Amazon Transcribe:
- Automated speech recognition (ASR) service used to transform the video audio into textual content transcripts
- Offers the muse for analyzing the video content material and figuring out participant particulars
- Permits for correct speech-to-text conversion, even in noisy sports activities environments
The next diagram illustrates the answer structure.
SkillShow AWS Structure Diagram
The architectural move is as follows:
- The approved person uploads a .csv file containing roster info (equivalent to jersey colour, quantity, participant identify, and college) and the video footage of gamers.
- A Lambda perform is triggered by the add of the video.
- The auto-transcript Lambda perform makes use of Amazon Transcribe to generate a timestamped transcript of what’s mentioned within the enter video.
- The transcript is uploaded to the output S3 bucket underneath
transcripts/for additional use. - The approved person can invoke the auto-clipper Lambda perform with an AWS Command Line Interface (AWS CLI) command.
- The perform parses the transcript towards participant info from the roster.
- When figuring out gamers, the perform clips movies primarily based on a specified key phrase (in SkillShow’s case, it was “Subsequent”) and uploads them to the output S3 bucket underneath
segments/.
By utilizing this suite of AWS providers, SkillShow was in a position to construct a scalable, cost-effective, and extremely automated video processing answer that addressed their key operational challenges. The cloud-based structure gives the flexibleness and scalability required to deal with their rising knowledge volumes and evolving enterprise wants.
Instance processing workflow
Let’s discover an instance processing workflow. As proven within the following screenshots, we first add a participant roster .csv and video file to the enter bucket.
The auto-transcribe perform processes the audio.
The auto-clipper perform segments the video primarily based on participant info.
Last clips are uploaded to the output bucket between two separate folders: a prefix of the enter video identify or Unnamed/ if the transcription was unclear or lacking the participant identify throughout the section.
Named movies will be seen within the first folder the place SkillShow’s present naming conference (jersey color_number_event video identify) is adopted for editors to obtain on demand.
Unnamed movies will be seen in an analogous naming conference, solely lacking the distinctive participant identify. Now, the editors solely need to assessment information on this folder and manually rename the file as an alternative of getting to do that for whole occasion movies.
Outcomes and advantages
After implementing this AWS powered answer, SkillShow reworked their video processing operations. The automated pipeline decreased video manufacturing time from 3 weeks to 24 hours per occasion, enabling sooner supply to athletes and scouts. A current occasion in Chicago showcased the system’s effectiveness. The automated pipeline processed 69 clips, precisely chopping and naming 64 of them—attaining a 93% success charge. This excessive accuracy demonstrates the answer’s capability to deal with real-world eventualities successfully. The system additionally proved adaptable, rapidly addressing preliminary challenges equivalent to colour naming inconsistencies.
The Northwest Indoor occasion additional illustrated the system’s scalability and flexibility. Right here, the automated course of dealt with a bigger quantity of roughly 270 clips, sustaining an estimated accuracy charge of over 90%. Notably, this occasion included batting observe footage, highlighting the answer’s adaptability to varied kinds of sports activities actions.
With this streamlined workflow, SkillShow has expanded its capability to course of a number of occasions concurrently, considerably enhancing its capability to serve youth sports activities leagues. The standardized output format and improved participant identification accuracy have enhanced the viewing expertise for athletes, coaches, and scouts alike. Though the time financial savings varies relying on particular occasion situations and filming methods, the system has demonstrated its potential to considerably cut back guide enhancing work. SkillShow continues to refine the method, rigorously balancing automation with high quality management to offer optimum outcomes throughout numerous occasion varieties. These enhancements positioned SkillShow to fulfill the rising demand for youth sports activities video content material whereas sustaining constant high quality throughout all occasions.
Conclusion
This answer demonstrates how AWS ML providers can rework resource-intensive video processing workflows into environment friendly, automated programs. By combining the scalable storage of Amazon S3, serverless computing with Lambda, and the speech recognition capabilities of Amazon Transcribe, organizations can dramatically cut back processing occasions and operational prices. As a pacesetter in automated sports activities video manufacturing, SkillShow has pioneered this strategy for youth sports activities whereas demonstrating its adaptability to varied content material varieties, from academic movies to company coaching. They’re already exploring extra synthetic intelligence and machine studying (AI/ML) capabilities for automated spotlight technology, real-time processing for reside occasions, and deeper integration with sports activities leagues and organizations.
For organizations trying to additional improve their video processing capabilities, Amazon Bedrock Knowledge Automation gives extra potentialities. Amazon Bedrock Knowledge Automation can streamline the technology of worthwhile insights from unstructured, multimodal content material equivalent to paperwork, photos, audio, and movies. This totally managed functionality may doubtlessly be built-in into workflows much like SkillShow’s, providing options equivalent to automated video summaries, content material moderation, and customized extraction of related info from video content material. Moreover, Amazon Bedrock Knowledge Automation can generate customized insights from audio, together with summaries and sentiment evaluation, offering even deeper understanding of spoken content material in sports activities movies.
SkillShow’s success highlights the broader potential of cloud-based video processing. As demand for video content material continues to develop throughout industries, organizations can use AWS ML providers to automate their workflows, cut back guide effort, and give attention to delivering worth to their clients somewhat than managing complicated enhancing operations.
Are you interested by implementing related automated video processing workflows on your group? Contact SkillShow to learn the way their pipeline constructed with AWS providers can rework your content material manufacturing course of.
In regards to the Authors
Ragib Ahsan is a Companion Options Architect at Amazon Internet Companies (AWS), the place he helps organizations construct and implement AI/ML options. Specializing in pc imaginative and prescient, he works with AWS companions to create sensible purposes utilizing cloud applied sciences. Ahsan is especially obsessed with serverless structure and its function in making options extra accessible and environment friendly.
Tom Koerick is the proprietor and CEO of SkillShow, a sports activities media community firm that has been filming youth sporting occasions nationwide since 2001. A former skilled baseball participant turned entrepreneur, Tom develops video options for occasion organizers and households within the youth sports activities business. His focus consists of school recruiting, social media sharing, and B2B providers that present added worth and income technology alternatives in youth sports activities.
