10 Steps to Outsource Video Annotation
Video Annotation – An Overview
An annotation, according to collinsdictionary.com, is a note that is added to a text or diagram, often in order to explain it.
An answer on Quora defines video annotation as ‘…the process of adding tags and labels to the video clips. It is used to help the computer vision-based models easily detect and identify the objects.’
Video annotation, in other, simple terms, is the process of adding on information to a video which would enable a software or computer to understand the data that a video represents.
Video is a complex set of data that contains information in many forms. From a contextual understanding of all of them together, a human being, through her intelligence, makes sense of this data. In its raw form, however, a machine is unable to make sense of it like a human being would. If we have to make it meaningful for a computer to understand, we have to enrich the information in the raw video data with annotations, or tags, so that it can be understood.
One way of looking at video as data is as a sequence of images, each a millisecond after the previous, creating a continuous motion to the naked eye, but identifiable as distinct images by a computer. The movement creates an added dimension to the data already contained in the image.
When we watch the recording of a football game, a human being intuitively understands that players wearing shirts of the same color constitute one team and players wearing shirts of a different color are the opposing team. However, for a computer, at the start, the shirt color is only raw data, and has no contextual reference. To enable it to understand this concept, an annotation has to be done that puts a name or a tag to the two different shirt colors so that a difference is created. Thereafter, through programming, the computer can be taught to treat the two colors as two different entities. Again, taking a simple illustration, if it was a video game, the computer can be taught that shirts of one color shoot at the goal on one end of the park while the other color shoots at the other end. This is what video annotation solutions do for us.
Why is this needed?
Artificial Intelligence (AI) and Machine Learning (ML) are perhaps the most important reason why video annotation is needed. In ML, through annotated data, a computer is taught to make sense of a video. Once it has been fed enough data and provided enough training, it can operate independently. In other words, it will reach a point where for the next video that it is fed, based on the training it has received, it will be able to interpret it in the same manner. For example, if an AI program is being readied to identify armed intrusions, one of the trainings that might be given is to identify a gun and the various shapes and sizes it could come in. Once the training is complete, the AI program can issue alerts based on which guards can take action, instead of having to manually go through footage beamed by hundreds of live surveillance cameras.
oWorkers has been providing data and video annotation services to clients in a variety of industries like Autonomous Vehicles, Medical AI, Satellite & Aerial imagery, Sports, Retail, Augmented Reality, Insurance, CCTV & Security, Robotics and Agriculture.
The 10 Steps to Outsource Video Annotation
In today’s day and age, outsourcing of video annotation services is almost a given. In fact, it is so much the norm that one may need to justify the case where one is NOT outsourcing rather than when one is.
Outlined here is a sequence of steps that, executed correctly will, hopefully, enable you to get the full benefit of video annotation and realise the purpose for which it was initiated. You must remember that equipping your vendor to do a good job is, eventually, in your own interest.
Identify Your Need
It is great that you think that the right thing to do is to outsource video annotation. Most of the world would agree with you.
But there is a much more basic question you should answer first. Do you have a need for video annotation solutions? In other words, why do you want video data to be annotated? What do you hope to achieve by doing so? Whether to do it inhouse or outsource it is, in a way, only a matter of detail once it has been established that there is a need for it in your business.
Of course, there are many smart vendors who will be able to tell you why video data should be annotated, because it is this business interest for more businesses to have this need. But they cannot answer on behalf of your business. Only you can. While vendors will willingly tell you how the video annotation process should be run and the benefits it will deliver, you should understand that the final ownership lies with you. Eventually it is an input into your business and needs to be driven by you and your business needs.
Will it solve a problem? Will it add value? What will be the criteria for success or failure?
Establish Selection Criteria
There are a host of factors that should be considered while evaluating potential partners, like technology capability, resourcing strength, multilingual coverage, financial stability, management support, etc.. All these parameters eventually manifest themselves, one way or another, in one of these three parameters in any delivery organisation, and hence should be accorded pride of place in the evaluation criteria:
Domain and function capability and the ability to deliver required quality and accuracy
oWorkers has been a preferred partner of choice for video annotation services for leading organisations, including several unicorn marketplaces, for over eight years. Many of our clients are referenceable. Through our three global delivery centers, we cater to over 22 global languages.
Ability to deliver the right quality within required timelines
With 24×7 operations, oWorkers is able to meet the most stringent timelines on client projects, in many cases delivering overnight. With clients from around the globe, we are also able to leverage time differences to expedite delivery.
Price point at which the capabilities are being made available and its reasonableness
oWorkers offers a choice of per unit or per hour pricing. Clients can choose. Our clients, especially from the US and Western Europe, have reported savings of 80% after outsourcing to oWorkers.
Put The Word Out, Invite Interest
Hopefully the first step has been taken, and it has been established that your business has a need for video annotation services. It has, perhaps, also been established at the same time that it is a time and resource consuming activity which you do not have the resources for. Hence, it has been agreed to outsource video annotation.
The next logical step is to start evaluating potential vendors. However, this does not happen automatically as we live in a world of information asymmetry. In other words, not everyone has access to all material information. Hence, at this stage, you need to ensure that your requirement is known reasonably well in the community from which will emerge the partner that will take over this task for you.
The Request for Proposal (RFP) process is a well-established process for B2B engagements. Whether you follow the RFP process or not, it is certainly recommended that you follow its discipline in some shape or form.
What this means is that you put the word out in such a way that you give out relevant information about the work that you seek to outsource, without divulging confidential information. You could provide information like:
- Brief description
- Volume of work
- Timelines expected
- Technology/ Tools to be used
- Manpower qualifications
How does it help?
It helps you by limiting responses to genuine parties who have an interest in doing the work that is available. Without information, there could be a slew of applications on the basis of incorrect assumptions, that you will spend time in reviewing and eventually eliminating because the work is not what they thought it was.
Identify Top Few, or Restart Process
The previous step would, hopefully, have resulted in a reasonable number of responses for video annotation solutions to your RFP or advertisement or whichever method you adopted to get the word out regarding your requirement.
At this stage, you need to put yourself in a position from where you are able to get into detailed discussions with a few suppliers out of which you are able to select a partner.
If you have received a large number of applications, you might want to evaluate the submissions and select down to a few, say two or three, that seem most appropriate.
If you have received a handful, say two or three, that seem relevant, you can avoid the earlier step of evaluation and retain them for a detailed discussion later.
But if you have either not received responses or received responses that are not relevant for your work, then you would need to go back to the drawing board and examine why that was the case.
Did you not define your requirements accurately?
Did you put out expectations that were unreasonable?
Detailed discussion, Final Selection and Contract
This is an obvious next step. You will now be in a position to hold detailed discussions and both parties will be required to share information, ordinarily preceded by the signing of an NDA (Non Disclosure Agreement). The vendors will tell you why they are the best to outsource video annotation to.
Specifics will be discussed at this stage, like pricing, like volumes, like other circumstances that could have an impact on the engagement.
Eventually it will lead to you identifying the most suitable partner and offering them a contract. However, at this stage, or even later, it will be wise to keep the other shortlisted parties interested in some way so that you have a Plan B if and when it may be needed. Selecting a vendor in a B2B relationship is a resource consumptive task. As it has downstream impact on the business that is selecting a partner, it is done carefully and deliberately. One does not want to be in a situation where the engagement breaks down soon after the process of selection has been completed, as it will mean going through it all over again. Hence keeping a second vendor warm is desirable.
The formal agreement is executed based on mutually acceptable terms and conditions.
If a Trial Run has been agreed this would be the time to initiate it. The contract would define the success criteria for this step and the rights of the client and vendor based on various outcomes possible.
If a vendor is not involved, a Trial Run is still a good idea before the organisation fully commits to it, hires resources, invests in technology, and everything else it entails. A sample of representative transactions would be carried out in this phase, enabling a further evaluation of the outcome and benefits, and enabling tweaking the process to enhance effectiveness, where required.
Developing and Implementing the Project Plan
Once it has been established that the organisation is ready to commit to it, a detailed Project Plan for the video annotation solutions will be developed, for both the outsourcer and the provider. This could also be done prior to commencement of the trial run.
The Project Plan will establish timelines for various tasks, define responsibilities, dependencies, checkpoints, controls mechanisms and other variables. In other words, what it means is that the work envisaged in the project is now being mainstreamed. The ‘Project Team’ or ‘Senior Management’ have shepherded the process so far, but henceforth will become a part of the responsibility of various teams based on the task and their role. Thus, the Project Plan defines ownership for various workstreams. Once this is activated, it is ‘all systems go.’
The Project Plan will take the initiative to the point of ‘steady state,’ the state where it should be operating hereafter, till forever, or till something happens that causes it to change, after having gone through a learning and improvement cycle earlier on.
When you outsource video annotation, once the Project Plan has been implemented and agreed, there will be various work streams that will be initiated. One of the key ones is technology. The vendor will need to ensure that the requisite technologies are available and ready to begin work for the new client. If client systems are needed, a handshake will be established.
Since we operate from the Eurozone, GDPR compliance is not a matter of choice. It is mandatory. We operate on secure technologies and protocols with ISO certifications (27001 :2013 & 9001:2015). We also offer physical segregation of workspaces where required.
Team Identification and Training
The other major workstream is ‘people.’ The vendor will need to identify the team that will support activities under this contract. Fresh resources will be hired as needed.
As subject matter experts, at this stage, are with the outsourcer, there will be an initial phase of ‘knowledge transfer’ or training during which the vendor staff will be made aware of the processing requirements.
Future training requirements, generally, are handled by the partner without the client staff having to go back to train more and more new hires.
oWorkers operates with employees, not outsourced staff or contractors. This provides it with the flexibility to deploy staff based on need. It has consistently been rated 4.6 and above out of 5 on Glassdoor by satisfied employees. It is deeply connected in the communities it operates in, pays social taxes for its employees and remains a desirable job for most people in the catchment area. With its deep connections, oWorkers gives you the flexibility to ramp up and down by 100 people in 48 hours, despite our strong screening processes based on Education, IQ, Language and Experience.
Ramp, Then Full Steam Ahead!
Eventually, the real work starts. In many cases, when you outsource video annotation, there is a gradual sequence of handling larger volumes each passing day and week as more resources become trained and available, and the vendor gets confidence in the new project. The work gradually ramps up to handle the agreed volumes over a period of time. This is also provided for in the contract and project plan in most cases.
oWorkers is a BPO services company with over eight years of experience and is led by a team with over twenty years of hands-on experience in the outsourcing industry. Being a pure player in data outsourcing services, Oworkers is in a position to provide comprehensive support to clients seeking to outsource video annotation. Our three global processing centers, in the most preferred regions of the world for outsourcing, can provide business continuity to your projects.