def predict_inv_unite(prob, anlg, tran, d):
b4 = prob.matrix[anlg.get("value")[3]]
b5 = prob.matrix[anlg.get("value")[4]]
pred_data = []
for ii, opt in enumerate(prob.options):
print(prob.name, anlg.get("name"), tran.get("name"), ii)
# prediction = transform.inv_unite(b4, b5, opt)
# score, _, _ = jaccard.jaccard_coef(opt, prediction)
b4_new, _, _, _, _ = transform.apply_binary_transformation(
b5, opt, transform.get_tran("unite"), imgC=b4)
score, _, _ = jaccard.jaccard_coef(b4_new, b4)
b5_score, _, _, _, _, _ = asymmetric_jaccard.asymmetric_jaccard_coef(
b5, opt)
opt_score, _, _, _, _, _ = asymmetric_jaccard.asymmetric_jaccard_coef(
opt, b5)
if max(b5_score, opt_score) > 0.85:
score = 0
pred_data.append({
**d, "optn": ii + 1,
"optn_score": score,
"mato_score": (d.get("mat_score") + score) / 2,
"pred": opt
})
return pred_data
def predict_preserving_subtract_diff(prob, anlg, tran, d):
best_diff_to_b2_x = d.get("diff_to_b2_x")
best_diff_to_b2_y = d.get("diff_to_b2_y")
best_diff_to_b3_x = d.get("diff_to_b3_x")
best_diff_to_b3_y = d.get("diff_to_b3_y")
best_diff = d.get("diff")
b4 = prob.matrix[anlg.get("value")[3]]
b5 = prob.matrix[anlg.get("value")[4]]
b2_ref = prob.matrix_ref[anlg.get("value")[0]]
prediction, best_diff_to_prediction_x, best_diff_to_prediction_y = transform.subtract_diff(
b5,
best_diff_to_b2_x,
best_diff_to_b2_y,
best_diff,
b2_ref,
coords=True)
pred_data = []
for ii, opt in enumerate(prob.options):
print(prob.name, anlg.get("name"), tran.get("name"), ii)
b5_score, diff_to_opt_x, diff_to_opt_y, diff_to_b5_x, diff_to_b5_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(opt, b5)
opt_to_b5_x = (-diff_to_opt_x) - (-diff_to_b5_x)
opt_to_b5_y = (-diff_to_opt_y) - (-diff_to_b5_y)
b3_to_b2_x = (-best_diff_to_b3_x) - (-best_diff_to_b2_x)
b3_to_b2_y = (-best_diff_to_b3_y) - (-best_diff_to_b2_y)
if abs(opt_to_b5_x - b3_to_b2_x) > 2 or abs(opt_to_b5_y -
b3_to_b2_y) > 2:
b5_score, diff = asymmetric_jaccard.asymmetric_jaccard_coef_pos_fixed(
opt, b5, b3_to_b2_x, b3_to_b2_y)
diff_score, _, _ = jaccard.jaccard_coef(diff, best_diff)
opt_score, _, _ = jaccard.jaccard_coef(prediction, opt)
b4_b5_aj = asymmetric_jaccard.asymmetric_jaccard_coef(b4, b5)
b4_opt_aj = asymmetric_jaccard.asymmetric_jaccard_coef(b4, opt)
preserving_score = min(b4_b5_aj[0], b4_opt_aj[0])
score = (diff_score + b5_score + diff_score + preserving_score) / 4
pred_data.append({
**d, "optn": ii + 1,
"optn_score": score,
"mato_score": (d.get("mat_score") + score) / 2,
"pred": prediction
})
return pred_data
def run_prob_anlg_tran_3x2_and_3x2(prob, anlg, tran):
b1_to_b2_x = None
b1_to_b2_y = None
diff_to_b3_x = None
diff_to_b3_y = None
diff_to_b2_x = None
diff_to_b2_y = None
diff = None
b1 = prob.matrix[anlg.get("value")[0]]
b2 = prob.matrix[anlg.get("value")[1]]
b3 = prob.matrix[anlg.get("value")[2]]
if "inv_unite" == tran.get("name"):
b1_new, _, _, _, _ = transform.apply_binary_transformation(
b2, b3, transform.get_tran("unite"), imgC=b1)
score, _, _ = jaccard.jaccard_coef(b1_new, b1)
elif "preserving_subtract_diff" == tran.get("name"):
b2_aj = asymmetric_jaccard.asymmetric_jaccard_coef(b1, b2)
b3_aj = asymmetric_jaccard.asymmetric_jaccard_coef(b1, b3)
preserving_score = min(b2_aj[0], b3_aj[0])
score, diff_to_b3_x, diff_to_b3_y, diff_to_b2_x, diff_to_b2_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(b3, b2)
if preserving_score < 0.85:
score = 0
else:
b1_b2_t, b1_to_b2_x, b1_to_b2_y, _, _ = transform.apply_binary_transformation(
b1, b2, tran, imgC=b3)
score, _, _ = jaccard.jaccard_coef(b1_b2_t, b3)
if "inv_unite" == tran.get("name"):
b2_score, _, _, _, _, _ = asymmetric_jaccard.asymmetric_jaccard_coef(
b2, b3)
b3_score, _, _, _, _, _ = asymmetric_jaccard.asymmetric_jaccard_coef(
b3, b2)
if max(b2_score, b3_score) > 0.9:
score = 0
if "unite" == tran.get("name") or "shadow_mask_unite" == tran.get("name"):
b1_score, _, _, _, _, _ = asymmetric_jaccard.asymmetric_jaccard_coef(
b1, b2)
b2_score, _, _, _, _, _ = asymmetric_jaccard.asymmetric_jaccard_coef(
b2, b1)
if max(b1_score,
b2_score) > 0.9: # if b1 is almost a subset of b2 or vice versa
score = 0
prob_anlg_tran_d = assemble_prob_anlg_tran_d(prob,
anlg,
tran,
score,
b1_to_b2_x=b1_to_b2_x,
b1_to_b2_y=b1_to_b2_y,
diff_to_b3_x=diff_to_b3_x,
diff_to_b3_y=diff_to_b3_y,
diff_to_b2_x=diff_to_b2_x,
diff_to_b2_y=diff_to_b2_y,
diff=diff)
return prob_anlg_tran_d
def predict_subtract_diff(prob, anlg, tran, d):
best_diff_to_u1_x = d.get("diff_to_u1_x")
best_diff_to_u1_y = d.get("diff_to_u1_y")
best_diff_to_u2_x = d.get("diff_to_u2_x")
best_diff_to_u2_y = d.get("diff_to_u2_y")
best_diff = d.get("diff")
u3 = prob.matrix[anlg.get("value")[2]]
u1_ref = prob.matrix_ref[anlg.get("value")[0]]
prediction, best_diff_to_prediction_x, best_diff_to_prediction_y = transform.subtract_diff(
u3,
best_diff_to_u1_x,
best_diff_to_u1_y,
best_diff,
u1_ref,
coords=True)
pred_data = []
for ii, opt in enumerate(prob.options):
print(prob.name, anlg.get("name"), tran.get("name"), ii)
u3_score, diff_to_opt_x, diff_to_opt_y, diff_to_u3_x, diff_to_u3_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(opt, u3)
opt_to_u3_x = (-diff_to_opt_x) - (-diff_to_u3_x)
opt_to_u3_y = (-diff_to_opt_y) - (-diff_to_u3_y)
u2_to_u1_x = (-best_diff_to_u2_x) - (-best_diff_to_u1_x)
u2_to_u1_y = (-best_diff_to_u2_y) - (-best_diff_to_u1_y)
if abs(opt_to_u3_x - u2_to_u1_x) > 2 or abs(opt_to_u3_y -
u2_to_u1_y) > 2:
u3_score, diff = asymmetric_jaccard.asymmetric_jaccard_coef_pos_fixed(
opt, u3, u2_to_u1_x, u2_to_u1_y)
diff_score, _, _ = jaccard.jaccard_coef(diff, best_diff)
opt_score, _, _ = jaccard.jaccard_coef(prediction, opt)
# prediction_to_opt_x = -to_pred_x + to_opt_x
# prediction_to_opt_y = -to_pred_y + to_opt_y
#
# best_diff_to_opt_x = best_diff_to_prediction_x + prediction_to_opt_x
# best_diff_to_opt_y = best_diff_to_prediction_y + prediction_to_opt_y
# best_diff_aligned, opt_aligned, _, _ = utils.align(best_diff, opt, best_diff_to_opt_x, best_diff_to_opt_y)
# u3_restored = utils.trim_binary_image(np.logical_or(best_diff_aligned, opt_aligned))
# u3_restored_score , _, _, _, _, _ = asymmetric_jaccard.asymmetric_jaccard_coef(u3, u3_restored)
score = (diff_score + u3_score + opt_score) / 3
pred_data.append({
**d, "optn": ii + 1,
"optn_score": score,
"mato_score": (d.get("mat_score") + score) / 2,
"pred": prediction
})
return pred_data
def evaluate_duplicate(u1, u2):
scores = []
u1_to_u2_xs = []
u1_to_u2_ys = []
current = u2.copy()
current_to_u2_x = 0
current_to_u2_y = 0
u1_tr = utils.trim_binary_image(u1)
while current.sum():
old_score_tmp, old_diff_tmp_to_u1_x, old_diff_tmp_to_u1_y, old_diff_tmp_to_current_x, old_diff_tmp_to_current_y, old_diff_tmp = \
asymmetric_jaccard.asymmetric_jaccard_coef(u1, current)
score_tmp, diff_tmp_to_u1_x, diff_tmp_to_u1_y, diff_tmp_to_current_x, diff_tmp_to_current_y, diff_tmp = \
soft_jaccard.soft_jaccard(u1, current, asymmetric = True)
if score_tmp < 0.9:
break
scores.append(score_tmp)
u1_to_current_x = (-diff_tmp_to_u1_x) - (-diff_tmp_to_current_x)
u1_to_current_y = (-diff_tmp_to_u1_y) - (-diff_tmp_to_current_y)
u1_to_u2_x = u1_to_current_x + current_to_u2_x
u1_to_u2_y = u1_to_current_y + current_to_u2_y
u1_to_u2_xs.append(u1_to_u2_x)
u1_to_u2_ys.append(u1_to_u2_y)
current = diff_tmp
current_to_u2_x = diff_tmp_to_current_x + current_to_u2_x
current_to_u2_y = diff_tmp_to_current_y + current_to_u2_y
if 1 >= len(scores):
mat_score = 0
u1_to_u2_locs = []
else:
mat_score = np.mean(scores)
u1_to_u2_xs = ((np.array(u1_to_u2_xs) - min(u1_to_u2_xs)) /
u1_tr.shape[1]).tolist()
u1_to_u2_ys = ((np.array(u1_to_u2_ys) - min(u1_to_u2_ys)) /
u1_tr.shape[0]).tolist()
u1_to_u2_locs = np.array(list(zip(u1_to_u2_xs, u1_to_u2_ys)))
stub = utils.make_stub(u1_to_u2_locs)
return mat_score, stub
def predict_add_diff(prob, anlg, tran, d):
best_diff_to_u1_x = d.get("diff_to_u1_x")
best_diff_to_u1_y = d.get("diff_to_u1_y")
best_diff_to_u2_x = d.get("diff_to_u2_x")
best_diff_to_u2_y = d.get("diff_to_u2_y")
best_diff = d.get("diff")
u3 = prob.matrix[anlg.get("value")[2]]
u1_ref = prob.matrix_ref[anlg.get("value")[0]]
prediction = transform.add_diff(u3, best_diff_to_u1_x, best_diff_to_u1_y,
best_diff, u1_ref)
pred_data = []
for ii, opt in enumerate(prob.options):
print(prob.name, anlg.get("name"), tran.get("name"), ii)
u3_score, diff_to_u3_x, diff_to_u3_y, diff_to_opt_x, diff_to_opt_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(u3, opt)
u3_to_opt_x = (-diff_to_u3_x) - (-diff_to_opt_x)
u3_to_opt_y = (-diff_to_u3_y) - (-diff_to_opt_y)
u1_to_u2_x = (-best_diff_to_u1_x) - (-best_diff_to_u2_x)
u1_to_u2_y = (-best_diff_to_u1_y) - (-best_diff_to_u2_y)
if abs(u3_to_opt_x - u1_to_u2_x) > 2 or abs(u3_to_opt_y -
u1_to_u2_y) > 2:
u3_score, diff = asymmetric_jaccard.asymmetric_jaccard_coef_pos_fixed(
u3, opt, u1_to_u2_x, u1_to_u2_y)
diff_score, _, _ = jaccard.jaccard_coef(diff, best_diff)
opt_score, _, _ = jaccard.jaccard_coef(opt, prediction)
score = (diff_score + opt_score + u3_score) / 3
pred_data.append({
**d, "optn": ii + 1,
"optn_score": score,
"mato_score": (d.get("mat_score") + score) / 2,
"pred": prediction
})
return pred_data
def run_prob_anlg_tran_2x2(prob, anlg, tran):
u1 = prob.matrix[anlg.get("value")[0]]
u2 = prob.matrix[anlg.get("value")[1]]
diff_to_u1_x = None
diff_to_u1_y = None
diff_to_u2_x = None
diff_to_u2_y = None
diff = None
copies_to_u1_x = None
copies_to_u1_y = None
u1_coms_x = None
u1_coms_y = None
u2_coms_x = None
u2_coms_y = None
if "add_diff" == tran.get("name"):
score, diff_to_u1_x, diff_to_u1_y, diff_to_u2_x, diff_to_u2_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(u1, u2)
elif "subtract_diff" == tran.get("name"):
score, diff_to_u2_x, diff_to_u2_y, diff_to_u1_x, diff_to_u1_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(u2, u1)
elif "xor_diff" == tran.get("name"):
score, u1_to_u2_x, u1_to_u2_y = jaccard.jaccard_coef(u1, u2)
score = 1 - score
u1_aligned, u2_aligned, aligned_to_u2_x, aligned_to_u2_y = utils.align(
u1, u2, u1_to_u2_x, u1_to_u2_y)
diff = utils.erase_noise_point(np.logical_xor(u1_aligned, u2_aligned),
4)
diff_to_u2_x = int(aligned_to_u2_x)
diff_to_u2_y = int(aligned_to_u2_y)
diff_to_u1_x = int(diff_to_u2_x - u1_to_u2_x)
diff_to_u1_y = int(diff_to_u2_y - u1_to_u2_y)
elif "upscale_to" == tran.get("name"):
u1_upscaled = transform.upscale_to(u1, u2)
score, _, _ = jaccard.jaccard_coef(u2, u1_upscaled)
elif "duplicate" == tran.get("name"):
scores = []
u1_to_u2_xs = []
u1_to_u2_ys = []
current = u2.copy()
current_to_u2_x = 0
current_to_u2_y = 0
while current.sum():
score_tmp, diff_tmp_to_u1_x, diff_tmp_to_u1_y, diff_tmp_to_current_x, diff_tmp_to_current_y, _ = \
asymmetric_jaccard.asymmetric_jaccard_coef(u1, current)
if score_tmp < 0.6:
break
scores.append(score_tmp)
u1_to_current_x = (-diff_tmp_to_u1_x) - (-diff_tmp_to_current_x)
u1_to_current_y = (-diff_tmp_to_u1_y) - (-diff_tmp_to_current_y)
u1_to_u2_x = u1_to_current_x + current_to_u2_x
u1_to_u2_y = u1_to_current_y + current_to_u2_y
u1_to_u2_xs.append(u1_to_u2_x)
u1_to_u2_ys.append(u1_to_u2_y)
u1_aligned, current_aligned, aligned_to_current_x, aligned_to_current_y = utils.align(
u1, current, u1_to_current_x, u1_to_current_y)
current = utils.erase_noise_point(
np.logical_and(current_aligned, np.logical_not(u1_aligned)), 8)
current_to_u2_x = aligned_to_current_x + current_to_u2_x
current_to_u2_y = aligned_to_current_y + current_to_u2_y
if 1 >= len(scores):
score = 0
copies_to_u1_x = [0]
copies_to_u1_y = [0]
else:
score = min(scores)
copies_to_u1_x = (np.array(u1_to_u2_xs[1:]) -
u1_to_u2_xs[0]).tolist()
copies_to_u1_y = (np.array(u1_to_u2_ys[1:]) -
u1_to_u2_ys[0]).tolist()
elif "rearrange" == tran.get("name"):
score, u1_coms_x, u1_coms_y, u2_coms_x, u2_coms_y = transform.evaluate_rearrange(
u1, u2)
else:
u1_t = transform.apply_unary_transformation(u1, tran)
score, _, _ = jaccard.jaccard_coef(u1_t, u2)
if "mirror" == tran.get("name") or "mirror_rot_180" == tran.get("name"):
# if u1 or u2 is already symmetric, then we shouldn't use mirror tran.
u1_mirror_score, _, _ = jaccard.jaccard_coef(u1_t, u1)
u2_mirror = transform.apply_unary_transformation(u2, tran)
u2_mirror_score, _, _ = jaccard.jaccard_coef(u2_mirror, u2)
if max(u1_mirror_score, u2_mirror_score) > 0.9:
score = 0
prob_anlg_tran_d = assemble_prob_anlg_tran_d(prob,
anlg,
tran,
score,
diff_to_u1_x=diff_to_u1_x,
diff_to_u1_y=diff_to_u1_y,
diff_to_u2_x=diff_to_u2_x,
diff_to_u2_y=diff_to_u2_y,
diff=diff,
copies_to_u1_x=copies_to_u1_x,
copies_to_u1_y=copies_to_u1_y,
u1_coms_x=u1_coms_x,
u1_coms_y=u1_coms_y,
u2_coms_x=u2_coms_x,
u2_coms_y=u2_coms_y)
return prob_anlg_tran_d
def run_prob_anlg_tran(prob, anlg, tran):
"""
compute the result for a combination of a problem, an analogy and a transformation.
:param prob:
:param anlg:
:param tran:
:return:
"""
diff_to_u1_x = None
diff_to_u1_y = None
diff_to_u2_x = None
diff_to_u2_y = None
diff = None
b1_to_b2_x = None
b1_to_b2_y = None
if 3 == len(anlg.get("value")): # unary anlg and tran
u1 = prob.matrix[anlg.get("value")[0]]
u2 = prob.matrix[anlg.get("value")[1]]
print(prob.name, anlg.get("name"), tran.get("name"))
if "add_diff" == tran.get("name"):
score, diff_to_u1_x, diff_to_u1_y, diff_to_u2_x, diff_to_u2_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(u1, u2)
elif "subtract_diff" == tran.get("name"):
score, diff_to_u2_x, diff_to_u2_y, diff_to_u1_x, diff_to_u1_y, diff = \
asymmetric_jaccard.asymmetric_jaccard_coef(u2, u1)
else:
u1_t = transform.apply_unary_transformation(u1, tran)
score, _, _ = jaccard.jaccard_coef(u1_t, u2)
elif 5 == len(anlg.get("value")): # binary anlg and tran
b1 = prob.matrix[anlg.get("value")[0]]
b2 = prob.matrix[anlg.get("value")[1]]
b3 = prob.matrix[anlg.get("value")[2]]
b1_b2_t, b1_to_b2_x, b1_to_b2_y, _, _ = transform.apply_binary_transformation(
b1, b2, tran)
score, _, _ = jaccard.jaccard_coef(b1_b2_t, b3)
else:
raise Exception("Ryan!")
return {
"prob_name": prob.name,
"anlg_name": anlg.get("name"),
"tran_name": tran.get("name"),
"pat_score": score, # pat = prob + anlg + tran
"prob_ansr": prob.answer,
"prob_type": prob.type,
"anlg_type": anlg.get("type"),
"tran_type": tran.get("type"),
"diff_to_u1_x": diff_to_u1_x,
"diff_to_u1_y": diff_to_u1_y,
"diff_to_u2_x": diff_to_u2_x,
"diff_to_u2_y": diff_to_u2_y,
"diff": diff,
"b1_to_b2_x": b1_to_b2_x,
"b1_to_b2_y": b1_to_b2_y
}
