Build Your Own AI Project Worksheet
AI tip (Accuracy = correct ÷ total × 100): Steps: 1 choose a problem → 2 collect examples → 3 label them → 4 check and correct bad labels (clean the data) → 5 split into training and test sets → 6 train → 7 test with new examples → 8 measure accuracy → 9 improve the data and retest. Accuracy = correct ÷ total × 100. Wrong = total − correct. Expected right = accuracy ÷ 100 × total. Keep some examples apart for testing (for example one fifth); the rest are for training.
NameDateTime takenScore ___ / 40
- 1.🧪 Your fruit sorter mixes up oranges and lemons. What helps most? (a) more labelled examples (b) a new colour screen (c) fewer examples (d) a louder speaker
- 2.📦 Rohan collected 170 labelled bird songs for a bird-song app and keeps one fifth of them for testing. How many are left for training?
- 3.📋 Planning an AI project: which comes first — label each example or calculate percent right? (a) label each example (b) calculate percent right
- 4.📦 Saanvi collected 110 labelled leaf photos for a leaf identifier and keeps one fifth of them for testing. How many are left for training?
- 5.📋 Planning an AI project: which comes first — set aside test photos or test on new photos? (a) set aside test photos (b) test on new photos
- 6.🎯 Meera’s handwriting reader is 80% accurate. If it is tested on 40 new handwritten letters, how many should it get right?
- 7.⚖️ Two fruit sorter models were tested on the same 20 fruit photos. Model A got 19 right and Model B got 13 right. How many percentage points more accurate is Model A?
- 8.📋 Planning an AI project: which comes first — work out the accuracy or fix it and test again? (a) work out the accuracy (b) fix it and test again
- 9.⚖️ Two spam spotter models were tested on the same 50 messages. Model A got 39 right and Model B got 37 right. How many percentage points more accurate is Model A?
- 10.⚖️ Two bird-song app models were tested on the same 50 bird songs. Model A got 36 right and Model B got 32 right. How many percentage points more accurate is Model A?
- 11.🧪 You can plan a simple AI project with… (a) nothing at all (b) paper and examples (c) a secret password (d) only a big robot
- 12.🎯 Kabir’s fruit sorter was tested on 5 new fruit photos and got 5 right. What is its accuracy in percent?
- 13.📦 Rohan collected 170 labelled leaf photos for a leaf identifier and keeps one fifth of them for testing. How many are left for training?
- 14.❌ Tara’s leaf identifier was tested on 50 new leaf photos and got 48 right. How many did it get wrong?
- 15.📋 Planning an AI project: which comes first — tag each photo by type or train the model? (a) tag each photo by type (b) train the model
- 16.📋 Planning an AI project: which comes first — split train and test or improve data, retest? (a) split train and test (b) improve data, retest
- 17.🎯 Kabir’s recycling sorter was tested on 4 new waste photos and got 4 right. What is its accuracy in percent?
- 18.🎯 Saanvi’s handwriting reader is 50% accurate. If it is tested on 90 new handwritten letters, how many should it get right?
- 19.📋 Planning an AI project: which comes first — calculate percent right or tag each photo by type? (a) calculate percent right (b) tag each photo by type
- 20.📋 Planning an AI project: which comes first — fix it and test again or try unseen examples? (a) fix it and test again (b) try unseen examples
- 21.🎯 Saanvi’s fruit sorter was tested on 5 new fruit photos and got 3 right. What is its accuracy in percent?
- 22.🎯 Diya’s fruit sorter is 60% accurate. If it is tested on 70 new fruit photos, how many should it get right?
- 23.🧪 What is accuracy? (a) total − correct (b) total × 2 (c) correct + total (d) correct ÷ total × 100
- 24.📋 Planning an AI project: which comes first — set aside test photos or train the model? (a) set aside test photos (b) train the model
- 25.🎯 Nisha’s handwriting reader is 50% accurate. If it is tested on 20 new handwritten letters, how many should it get right?
- 26.📋 Planning an AI project: which comes first — recheck every label or test on new photos? (a) recheck every label (b) test on new photos
- 27.📦 Diya collected 150 labelled leaf photos for a leaf identifier and keeps one fifth of them for testing. How many are left for training?
- 28.🎯 Meera’s leaf identifier was tested on 4 new leaf photos and got 4 right. What is its accuracy in percent?
- 29.🧪 Why must you test a model on examples it has NOT seen? (a) to make it bigger (b) to see if it learned (c) to save paper (d) to make it faster
- 30.⚖️ Two handwriting reader models were tested on the same 25 handwritten letters. Model A got 23 right and Model B got 15 right. How many percentage points more accurate is Model A?
- 31.📋 Planning an AI project: which comes first — fix it and test again or work out the accuracy? (a) fix it and test again (b) work out the accuracy
- 32.📦 Meera collected 160 labelled messages for a spam spotter and keeps one fifth of them for testing. How many are left for training?
- 33.🎯 Vihaan’s bird-song app was tested on 5 new bird songs and got 3 right. What is its accuracy in percent?
- 34.📋 Planning an AI project: which comes first — train the model or improve data, retest? (a) train the model (b) improve data, retest
- 35.🎯 Vihaan’s leaf identifier is 50% accurate. If it is tested on 70 new leaf photos, how many should it get right?
- 36.📦 Kabir collected 90 labelled fruit photos for a fruit sorter and keeps one fifth of them for testing. How many are left for training?
- 37.🎯 Diya’s fruit sorter was tested on 10 new fruit photos and got 10 right. What is its accuracy in percent?
- 38.⚖️ Two bird-song app models were tested on the same 50 bird songs. Model A got 48 right and Model B got 42 right. How many percentage points more accurate is Model A?
- 39.📋 Planning an AI project: which comes first — improve data, retest or test on new photos? (a) improve data, retest (b) test on new photos
- 40.❌ Ishaan’s bird-song app was tested on 22 new bird songs and got 16 right. How many did it get wrong?
Answer key
- more labelled examples
- 136
- label each example
- 88
- set aside test photos
- 32
- 30
- work out the accuracy
- 4
- 8
- paper and examples
- 100
- 136
- 2
- tag each photo by type
- split train and test
- 100
- 45
- tag each photo by type
- try unseen examples
- 60
- 42
- correct ÷ total × 100
- set aside test photos
- 10
- recheck every label
- 120
- 100
- to see if it learned
- 32
- work out the accuracy
- 128
- 60
- train the model
- 35
- 72
- 100
- 12
- test on new photos
- 6
Free ai for students lessons, practice and worksheets at talentjr.in/ai-for-students