Computer Vision — professional stack.
For each requirement below, pick the option that fits your build — recommended first, then free and cheaper alternatives — or skip what your project doesn't need. Tap the info icon next to any requirement to see why it matters.
Image/Video Object Storage (Petabyte Ingestion)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsAWS S3 Free 5GB 12mo; R2 Free 10GB storage + 10M Class A ops + zero egress; S3 $0.023/GB/mo Standard, PUT $0.005/1k, GET $0.0004/1k, egress $0.09/GB; R2 $0.015/GB/mo store + $4.50/M writes, $0.36/M reads; Wasabi $6.99/TB/mo no egress; overage per GB billed
GPU Inference Compute (A10G / A100 / H100) - CV Detection & Segmentation
Where your code actually runs and serves requests. Picking the right host affects speed, scaling, and how much ops work you do.Pricing & free-tier limitsColab Free T4 ~12hrs session limited; Kaggle 30hrs P100/week; HF Free T4 CPU; RunPod A4000 $0.39/hr, A100 40GB $0.79/hr, A100 80GB $1.49/hr, H100 $3.98/hr; Modal $30/mo free credits then $0.000725/sec A10G ($2.61/hr) $0.0016/sec A100 ($5.76/hr billed per sec); overage per sec autoscale
CV Model Hosting & Serverless Inference (Replicate, Roboflow, HuggingFace)
Where your code actually runs and serves requests. Picking the right host affects speed, scaling, and how much ops work you do.Pricing & free-tier limitsHF Inference Free 30k tokens/day limited; Replicate Free $10 credits; Replicate $0.000725/sec A10G ($2.61/hr), $0.0016/sec A100 ($5.76/hr) per sec; Roboflow Free 1k/mo, Starter $249/mo 50k inf ($0.0049/inf overage), Growth $999/mo 250k; HF Endpoints CPU $0.60/hr, T4 $1.20/hr, A10G $1.81/hr; overage billed per sec
Image Optimization & Global CDN Delivery
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsCloudflare Free CDN unlimited BW; Images $5/100k stored/mo + $1/100k delivered; Cloudinary Free 25k trans + 25GB BW + 25GB storage, Growth $89/mo 100k trans 100GB BW; Imgix Free 1k master images 100GB BW, $10/mo + $3/1k extra images, $0.08/1000 image req; Bunny CDN $0.01/GB + $9/mo min; overage $0.02/GB
Vector Database for Image Embeddings (Visual Search - CLIP, DINOv2)
Persistent storage for your app’s data — users, products, orders. The single most important architectural decision for most projects.Pricing & free-tier limitsPinecone Free 100k vectors 1 index 2GB storage; Qdrant Free 1GB RAM ~1M 768d vectors 5GB disk; Pinecone Serverless Starter $70/mo includes 2M vectors + 100GB storage, storage $0.33/GB, reads $0.08/M 768d, writes $2/M; Qdrant $25/mo 1GB RAM 10GB disk 1B search ops, extra $0.20/GB; Supabase pgvector free tier; overage Pinecone $0.35/GB storage
Background Job Orchestration for Video Processing
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsInngest Free 50k steps 100 GB-hr executions; SQS Free 1M msgs/mo (first 12mo) then $0.40/M standard $0.50/M FIFO; Cloudflare Queues Free 1M ops then $0.40/M; QStash Free 500 msgs/day; Trigger.dev Free 10k runs, Pro $20/mo 50k runs + $0.0005/run overage; Inngest Pro $199/mo 500k steps then $0.00015/step; overage SQS $0.40/M
CV Job Queue & Task Scheduling
Runs slow work (emails, exports, payouts) in the background so users don’t wait. Keeps your app snappy.Pricing & free-tier limitsUpstash Free 10k cmds/day 256MB; Redis Cloud Free 30MB Essential, $5/mo 100MB, Pro $20/mo 1GB; SQS Free 1M msgs/mo $0.40/M after; Sidekiq Pro $299 one-time license + $179/yr upkeep; Upstash $10/mo 100k cmds/day 1GB; overage Upstash $0.20/100k cmds, Redis Cloud $0.15/100MB
Annotation Tool Hosting (Label Studio / CVAT)
Where your code actually runs and serves requests. Picking the right host affects speed, scaling, and how much ops work you do.Pricing & free-tier limitsLabel Studio OSS Free self-host infra only ~$20/mo EC2; CVAT.org Free OSS; Label Studio Cloud Starter Free 10k annotations 1GB; Team $99/mo 50k annos 10GB; CVAT.ai Free 10 tasks 3k jobs, Team $33/user/mo cap $300/mo 10k tasks; Labelbox Free 10k labels, Starter platform $750/mo min + $0.12/label bbox $0.25/seg; overage $0.08/label QA
Video Transcoding & Adaptive Streaming (Mux, Cloudinary, AWS Elemental)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsMux Free $20 credit ~500min SD; Cloudflare Stream Free trial 1000 min, then $5/1000 min stored + $1/1000 min delivered; Mux $0.04/min 1080p, $0.015/min SD, $0.08/min 4K encode, delivery $0.004/min; MediaConvert $0.0075/min SD $0.015/min HD $0.03/min 4K; Bunny Stream $0.005/GB encode + $0.01/GB delivery; overage Mux $0.02/min storage
Model Registry & Versioning (MLflow, W&B Registry, HuggingFace Hub)
Powers AI features — model access, embeddings, and inference. Costs scale with usage, so watch the meter.Pricing & free-tier limitsMLflow OSS Free only S3 ~$0.023/GB; W&B Free 100GB artifacts + 100hrs tracking 1 member; HF Hub Free unlimited public repos, Pro $9/mo private; DagsHub Free 10GB storage; W&B Team $20/user/mo 200GB extra 1k hrs; Databricks Managed MLflow $0.07/DBU + $0.40/hr managed; overage W&B $0.50/GB storage + $8/100hrs
Experiment Tracking & Training Observability (W&B, MLflow, TensorBoard)
Tells you when things break and why. You cannot fix what you cannot see — this is how you keep downtime short.Pricing & free-tier limitsW&B Free 100 tracked hrs + 100GB storage 1 member; MLflow OSS Free self-host EC2 $20/mo; TensorBoard.dev free deprecating; Comet Free 100 experiments; W&B Team $20/user/mo 200GB 1000hrs, Enterprise $50/user unlimited; Comet Pro $49/mo 100GB + 500 experiments; overage W&B $8 per 100 extra hrs + $0.50/GB
Training Infrastructure for CV Fine-tuning (GPU Cluster, Modal, Anyscale)
Powers AI features — model access, embeddings, and inference. Costs scale with usage, so watch the meter.Pricing & free-tier limitsKaggle Free 30hrs P100/week + 20GB persistent; Colab Free T4 limited; Modal $30 free credits/mo then A100 $1.10/hr, A10G $0.74/hr billed per sec min 1 sec; RunPod A100 40GB $0.79/hr spot $0.39/hr, H100 $3.98/hr spot $2.5/hr; CoreWeave A100 $1.49/hr H100 $4.19/hr; Anyscale A100 $1.50/hr + $1/hr mgmt; P4d $32.77/hr 8xA100 ($4.09/GPU/hr); overage per sec
Image Metadata Search & Indexing (Algolia, Typesense, Elasticsearch)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsAlgolia Free 10k records 10k search req/mo; Typesense Free 500MB 100k docs; Meilisearch OSS Free; Typesense $14/mo 1GB 500k docs 128GB bandwidth; Algolia Growth $250/mo 100k records 1M ops then $0.40/1k records $0.50/1k ops; Elastic Cloud $95/mo 1GB RAM 2vCPU; overage Algolia $0.60/1k ops + $0.40/1k records
Monitoring & Inference Latency Tracking (Prometheus, Grafana, Evidently, Langfuse)
Tells you when things break and why. You cannot fix what you cannot see — this is how you keep downtime short.Pricing & free-tier limitsGrafana Free 10k metrics 50GB logs 50GB traces 14d; Langfuse Free 50k observations/mo 1 member; Evidently OSS Free; Langfuse Hobby $39/mo 100k obs + $0.30/1k overage, Pro $199/mo 500k obs + $0.20/1k; Datadog APM Pro $31/host/mo + $1.70/10k spans traced; Prometheus self-host free; overage Grafana $0.40/1000 series $0.50/GB logs
Secrets Management & Model Key Vault
Stores API keys and credentials safely, separate from your code. Leaking secrets is one of the most common breaches.Pricing & free-tier limitsDoppler Free 5 users 100 secrets 25 config tokens; Infisical OSS Free unlimited self-host; Infisical Cloud Free 15k secret ops; Doppler $7/user/mo 500 secrets unlimited tokens; AWS Secrets Manager $0.40/secret/mo + $0.05/10k API calls $0.03 after; Vault Cloud HCP $0.03/hr cluster ~$22/mo; DOP $12/user/mo Enterprise; overage AWS $0.05/10k calls
CI/CD for Model Deployment (GitHub Actions, Docker, Model CI)
Automates testing and deploying your code. Saves enormous time and prevents “works on my machine” releases.Pricing & free-tier limitsGitHub Free 2000 Linux x64 min + 500MB packages 500 min ARM; CircleCI Free 6k credits (~1000 min); GitLab Free 400 min; GitHub Team $4/user 3k min, extra Linux $0.008/min, macOS $0.08/min, GPU $0.032/min (A100), Windows $0.016; CircleCI $15/mo 30k credits + $0.02/credit over; overage Actions $0.008/min Linux
Data Labeling Pipeline & QA Workflow (Scale, Labelbox, V7)
A piece of your stack you may or may not need, depending on scope. Pick the option that fits — or skip it if your project doesn’t require this capability yet.Pricing & free-tier limitsLabelbox QA Free integrated workflow; Scale AI Rapid self-serve $0.05/bbox + $0.02 QA per task + $0.15/seg; V7 Darwin Free 10 users 50k images; Label Studio Team $99/mo 50k annos; V7 Pro $399/mo 250k images + workflows; Labelbox Platform $750/mo min 10k labels included + $0.12/label overage; Scale Enterprise from $2k/mo; overage Scale $0.08/bbox seg $0.25
Content Moderation API (Hive, AWS Rekognition Moderation, Sightengine)
Manages the content and assets you publish without editing code. Essential for sites with frequent updates.Pricing & free-tier limitsSightengine Free 2k req/mo image video moderation; Hive Free 1k req; AWS Rekog Free 5k images/mo 12mo then image moderation $1/1k first 1M $0.80/1k next 9M $0.60/1k >10M; Video $0.10/min; Hive $1.50/1k images mod, adult $1/1k, video $2.50/1k sec; Sightengine $14/mo 10k ops $0.004/extra; overage AWS $0.001/image tiered
Container Registry & Orchestration (Docker Hub, ECR, Kubernetes)
A piece of your stack you may or may not need, depending on scope. Pick the option that fits — or skip it if your project doesn’t require this capability yet.Pricing & free-tier limitsGHCR Free 500MB storage 1GB transfer public; Docker Hub Free 1 private repo unlimited pulls 100 pulls/6hr; AWS ECR Free 500MB 12mo then $0.10/GB/mo storage + $0.09/GB egress; EKS Control Plane $0.10/hr ($72/mo) + EC2; GKE Standard $0.10/hr per cluster (~$72/mo); overage GHCR storage $0.25/GB $0.50/GB transfer; ECR $0.10/GB
Evaluation & Model Performance Testing (FiftyOne, Evidently, W&B Eval)
Powers AI features — model access, embeddings, and inference. Costs scale with usage, so watch the meter.Pricing & free-tier limitsFiftyOne OSS Free self-host only compute; FiftyOne Teams $250/mo includes 20TB data inspected 5 users; Evidently OSS Free 0 cost; Evidently Cloud Free 10k metrics; Teams $200/mo 50k metrics + $0.01/1k overage; W&B Weave Free 50k traces, Pro $200/mo 500k traces $0.30/1k extra; overage $0.02/GB scanned eval dataset
Feature Store for Visual Features / Embeddings Cache
Stores frequent results in fast memory so you serve less from the database. Big lever for speed and cost.Pricing & free-tier limitsFeast OSS Free only Redis $10/mo Upstash + S3; Hopsworks Free 3 users 15GB 10k features; Redis 30MB free; Tecton Team $2k/mo includes 10M feature reads + 1TB offline store, $0.05/1k extra online reads $0.10/1k offline; Hopsworks Pro $250/mo 50GB; Redis Cloud Pro $50/mo 1GB; overage Tecton $0.10/1k reads + $0.02/GB offline
Data Versioning & Lineage (DVC, LakeFS, Pachyderm)
A piece of your stack you may or may not need, depending on scope. Pick the option that fits — or skip it if your project doesn’t require this capability yet.Pricing & free-tier limitsDVC OSS Free only S3 cost $0.023/GB; DVC Studio Free 10GB versioning; Team $20/user/mo 100GB; LakeFS Free 10GB 50k objects; $40/mo 100GB + 500k objects + $0.10/GB overage; Pachyderm Community Free 1 pipeline 10GB; Enterprise $1k/mo 10 pipelines 1TB, $100/pipeline over; overage LakeFS $0.10/GB
Distributed Training & Hyperparameter Tuning (Ray, RunPod, CoreWeave)
A piece of your stack you may or may not need, depending on scope. Pick the option that fits — or skip it if your project doesn’t require this capability yet.Pricing & free-tier limitsRay OSS Free only EC2; W&B Sweeps Free 100 runs; Anyscale Free $25 credits then $1/hr cluster mgmt + $1.50/hr A100 node $6/hr 8xA100; RunPod Multi-GPU 2xA100 $1.58/hr, 4xA100 $3.16/hr, 8xA100 $12.64/hr; CoreWeave CKS $0.10/hr + GPU $1.49/hr A100; W&B Launch $0.50/hr queue; overage Anyscale $2/hr per worker
Real-time Video Analytics Pipeline (Kafka, Flink, DeepStream)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsConfluent Free $200 credit 30d + 10GB storage $0.11/GB/mo after; Upstash Kafka Free 10k msgs/day 256MB, $10/mo 100k/day 10GB; Redpanda Free 10GB; Confluent Standard $0.20/hr CKU ($144/mo per CKU) + 100GB $11/mo + $0.11/GB ingress/egress; Flink $0.40/hr per CU; MSK $0.30/hr broker; DeepStream SDK free, Enterprise Triton $1k/mo support; overage $0.115/GB + $0.40/hr CU
Face/Person Detection Privacy & Compliance (PII Redaction, GDPR, BIPA)
Keeps you secure and compliant with regulations. Required for enterprise customers and handling sensitive data.Pricing & free-tier limitsCompreFace OSS Free infra only ~$30/mo EC2 g4dn; AWS Rekog Face Free 5k faces/mo 12mo then $1/1k first 1M $0.80 next 9M $0.60 after; Collection $0.0005/face/mo storage; Private AI Free 100 ops/mo, Starter $50/mo 10k images ($0.005/img overage), Scale $500/mo 150k; face blur GPU processing $0.002/image; overage AWS $0.0008 per face
Audit Logs & WORM Storage for Vision Data (GDPR/CCPA Compliant Retention)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsS3 Object Lock free feature enabling WORM compliance mode; storage $0.023/GB Standard, $0.0125/GB IA, $0.004/GB Glacier, $0.00099/GB Deep Archive; CloudTrail first trail free management events, $2.00/100k data events; Axiom Free 500MB/day 30d ret, $25/mo 100GB ingest 30d $0.25/GB overage; Datadog Logs $0.10/GB ingest + $1.70/M indexed + 15d retention; WORM 7yr GDPR biometric retention required $0.004/GB/mo; overage $0.03/GB retrieval
Edge Deployment & ONNX Optimization (TensorRT, OpenVINO, OctoML)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsONNX Runtime OSS Free; TensorRT OSS Free; OpenVINO Free; Triton Inference Server OSS Free self-host; OctoML Free 10 model optimizations 1 user, Scale $500/mo 100 optims then $5/optim + $0.10/hr inference acceleration; Cloudflare Workers AI Free 10k neurons/day ~10M tok, Paid $5/mo includes 10M neurons + $0.011/1k neurons overage ($0.011/1k tokens); Cloudflare GPU $0.50/hr Workers; overage OctoML $5/model
API Gateway & Rate Limiting for CV APIs (Model API Management)
Controls and protects your APIs — quotas, abuse prevention, and firewalls. Important once you have real traffic or many clients.Pricing & free-tier limitsCloudflare Free 100k API requests/day WAF rate limit unlimited; AWS API GW Free 1M calls 12mo then HTTP API $1.00/M + $0.09/GB transfer, REST API $3.50/M + $0.09/GB; Kong Konnect Free 1M gateway calls/mo; Team $250/mo 10M calls, $0.20/1M overage; Zuplo Free 10k/day, Dev $20/mo 500k, Pro $300/mo 10M + $0.30/1M over; overage AWS $3.50/M
Uptime Monitoring & Incident Response (Model API Health)
Tells you when things break and why. You cannot fix what you cannot see — this is how you keep downtime short.Pricing & free-tier limitsBetterStack Free 10 monitors 3min checks + 10 on-call + status page; Grafana Cloud Free 10k metrics + OnCall Free 5 users alerts; UptimeRobot Free 50 monitors 5min; PagerDuty Free 5 users 10 incidents; BetterStack Startup $24/mo 50 monitors 30sec + 3 status pages; Team $99/mo 200 monitors 10sec; PagerDuty Pro $41/user/mo 500k events; overage BetterStack $0.10/monitor $0.20/1k on-call alerts
Error Tracking & Model Failure Logging (Sentry, Axiom)
Tells you when things break and why. You cannot fix what you cannot see — this is how you keep downtime short.Pricing & free-tier limitsSentry Free 5k errors 10k transactions replay limited 50 replays; Team $26/mo 50k errors 100k tx 500 replays; Business $80/mo 100k errors $0.0003/extra error $0.0002/tx; Axiom Free 0.5GB/day 30d, $25/mo 100GB 30d $0.25/GB overage; Highlight Free 500 sessions, Startup $50/mo 10k sessions $0.01 extra; overage Sentry $0.0004/error $0.0005/replay
How this professional computer vision checklist works.
Each requirement below is something a professional computer vision build typically needs. Pick one of the four researched options — recommended, free, cheaper or paid — add your own with "Other", or skip the requirement if your project doesn't need it. Nothing is mandatory; the plan on the right tracks what you've decided so nothing gets forgotten.
Your picks are saved in this browser automatically, so you can come back anytime. Options are researched per build level and refreshed as vendors change their plans — always verify details on the provider's page before committing.